Contactless WiFi Sensing and Monitoring

for Future Healthcare: Emerging Trends,

Challenges and Opportunities
Ge, Y., Taha, A., Shah, S. A., Dashtipour, K., Zhu, S., Cooper, J.
M., Abbasi, Q. & Imran, M.
Published PDF deposited in Coventry University’s Repository

Original citation:
Ge, Y, Taha, A, Shah, SA, Dashtipour, K, Zhu, S, Cooper, JM, Abbasi, Q & Imran, M
2022, 'Contactless WiFi Sensing and Monitoring for Future Healthcare: Emerging
Trends, Challenges and Opportunities', IEEE Reviews in Biomedical Engineering.
https://doi.org/10.1109/RBME.2022.3156810

DOI 10.1109/RBME.2022.3156810
ISSN 1937-3333
ESSN 1941-1189

Publisher: Institute of Electrical and Electronics Engineers

This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Contactless WiFi Sensing and Monitoring for Future Healthcare -

Emerging Trends, Challenges and Opportunities
Yao Ge, Ahmad Taha, Syed Aziz Shah, Kia Dashtipour, Shuyuan Zhu, Member, IEEE, Jonathan Cooper, Qammer
H. Abbasi, Senior Member, IEEE, Muhammad Ali Imran, Senior Member, IEEE,

Abstract—WiFi sensing has received recent and significant healthcare applications, demonstrating its relative advantages
interest from academia, industry, healthcare professionals, and over other monitoring systems such as sensing methods, such
other caregivers (including family members) as a potential as wearable sensors, camera-based imaging, and acoustic-based
mechanism to monitor our aging population at a distance without
deploying devices on users’ bodies. In particular, these methods solutions.
have the potential to detect critical events such as falls, sleep
disturbances, wandering behavior, respiratory disorders, and A. Comparison of WiFi RF sensing and other approaches
abnormal cardiac activity experienced by vulnerable people.
The interest in such WiFi-based sensing systems arises from Broadly, current sensing and monitoring systems can be
practical advantages including its ease of operation indoors as divided into those using contact-based sensors, including
well as ready compliance from monitored individuals. Unlike other wearables [1] and contactless systems [2], [3]. Besides wearable
sensing methods, such as wearables, camera-based imaging, and
acoustic-based solutions, WiFi technology is easy to implement devices, the contactless monitoring approaches can be divided
and unobtrusive. This paper reviews the current state-of-the-art into visual based sensing radio frequency (RF) signals based
research on collecting and analyzing channel state information sensing. RF signals at frequencies between 30 kHz & 300
extracted using ubiquitous WiFi signals, describing a range of GHz, comprise electromagnetic waves called radio waves (as
healthcare applications and identifying a series of open research are widely used in radar systems, including household and
challenges, including untapped areas of research and related
trends. This work aims to provide an overarching view in commercial behavior recognition [4]). Recently, the carrier
understanding the technology and discusses its use-cases from a frequency range of WiFi signals is from 2.4 GHz to 5.9 GHz,
perspective that considers hardware, advanced signal processing, which is covered by radio waves.
and data acquisition. Applications of wearable devices cover a wide range of
methods, including measurements of heartbeat and respiration
Index Terms—WiFi sensing, healthcare detection, machine rates, oxygen saturation level, electromyographic signals and
learning, deep learning many others [5], [6]. However, these sensors are expensive,
as it is necessary to provide a single device to each person
I. I NTRODUCTION being monitored. Moreover, the successful capture of the health
Sensing and monitoring systems for human healthcare information is dependant on the patient wearing the sensor
have become increasingly popular, driven in part through our or keeping it close to the body, which, if forgotten, can have
knowledge economy as well as the significant improvements in severe consequences in applications such as fall detection.
our longevity and living standards. In healthcare applications, There is also the challenge of the re-usability of wearable
such systems can provide individuals with the capability of equipment, resulting in widespread contact-transmission viruses,
long-term detection of daily activities and variations in vital such as COVID-19, if not appropriately disinfected. Generally
signs, all in the privacy of our homes. With simple, long- the adoption of the technology is problematic amongst some
term, and continuous health monitoring in the daily home of the most needy individuals, namely those who are old or
environment, it is possible to record the signs of illness and disabled.
physiological deterioration that cannot be detected during a In contactless sensing methods, camera-based sensing ap-
short formal clinical consultation. Such monitoring systems plications have proven their accuracy [7]. However, several
can also be combined with deep learning and can be used to disadvantages make it difficult, in some scenarios, to rely on
monitor behavior, including emotional states and mental well- such systems, which includes:
being. Such information can be integrated into smart homes to • System complexity and high cost due to computational
support our daily lives. In this study, we focus on a detailed requirements for multiple cameras to cover areas of
review which explores the application of WiFi sensing in such activity.
• Privacy concerns due to the capturing and storage of
Yao Ge, Ahmad Taha, Kia Dashtipour, Jonathan Cooper, Muhammad images, which unauthorized users can access in a low-
Ali Imran and Qammer H. Abbasi are with the James Watt
School of Engineering, University of Glasgow, Glasgow, UK. e- security system.
mail: (2288980g@student.gla.ac.uk; ahmad.taha@glasgow.ac.uk; Compared with wearable sensing technologies, ambient RF
kia.dashtipour@glasgow.ac.uk; jon.cooper@glasgow.ac.uk; muham-
mad.imran@glasgow.ac.uk; qammer.abbasi@glasgow.ac.uk). sensing has the advantage of reducing the risk of contact
Syed Aziz Shah is with the Centre for Intelligent Healthcare, Coventry transmission infections. Because it is capable of the contactless
University, Coventry, UK. e-mail: (azizshahics@yahoo.com). measurement of vital signatures and macro-health indicators in
Shuyuan Zhu is with the School of Information and Communication
Engineering, University of Electronic Science and Technology of China, non-line-of-sight (NLOS) environments. In hospitals, wireless
Chengdu, China. e-mail: (eezsy@uestc.edu.cn). systems can capture the signal signature of vital signs, such as

1
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

coughing, shortness of breath, fever, and aches [8]–[10]. Given review, providing a healthcare perspective to researchers. More
that these symptoms are closely linked to patient infections, specifically, the contributions are as follows:
WiFi sensing has the potential to detect illness. The comparison • Provide a detailed overview of the methodologies adopted
of WiFi sensing and other RF sensing technique is conducted in developing healthcare monitoring systems.
and discussed in Section II-B. • Classify healthcare related applications into different
categories, and then provide insights for distinct trends.
B. Specification of WiFi sensing in healthcare • Highlight challenges and their potential solutions that re-

At present, WiFi sensing research in healthcare is being quire further investigation for generalization of healthcare
mostly developed for use in non-hospital environments, driven applications based on WiFi sensing.
by two trends: vital sign detection and activity detection The paper is organized as follows: Section II introduces WiFi
(see Fig. 1). Vital sign detection system aims to monitor the sensing technical background and development in healthcare
movement of the lungs and heart in humans using WiFi signals range. Section III reviews different techniques applied in
to recognize the respiration and heartbeat rate, in real-time. WiFi sensing, including signal preprocessing techniques and
For activity detection task, alarms for critical events such as algorithms. Section IV concludes and analyzes the recent
falling, and other specific actions that can cause severe and fatal healthcare related applications in different fields, including
consequences to human-beings has been studied in academia, vitals detection, localization, large-scale and small scale activity
and industry [11]. recognition. While finally, Section V discusses the technical
Generally, such systems use WiFi devices alongside intel- and ethical challenges based on the recent researches. Then
ligent classification algorithms to monitor and predict human provides future perspectives associated with healthcare WiFi
subjects’ movements. In the same context, WiFi signals are sensing.
also used to report, over the internet, the activity status and/or
vitals of the monitored subjects to the medical specialist and II. R ELATED W ORK OF W I F I SENSING
families or carers. Beneficiaries of such valuable real-time
data and information are the internet of things (IoT) systems This section presents the existing research studies focused
[12]. For example, vital signs detection in a smart home can on the technical background of WiFi monitoring systems
help IoT systems adjust the temperature, humidity, and other with CSI and received signal strength indicator (RSSI) and
environmental factors automatically to improve the quality of descriptions of different tracks of WiFi sensing technology.
the user’s experience [13], [14]. At present, WiFi sensing has Meanwhile, human activity recognition based on other RF
been applied in the home. For example, Linksys sells a WiFi sensing technology is introduced.
router and provides a service called ”Linksys Aware,” which
enables WiFi devices to perceive the signals’ vibration around A. Technical background of WiFi sensing
the house. Although there have been numerous research studies
With the rapid advances in communication and network
conducted in this field, it is difficult to replace wearable and
technology, it is possible to assume the broad deployment of
visualized healthcare applications due to their high reliability
WiFi devices across society. Multiple-input multiple-output
and efficiency. However, as academia and industry continue to
(MIMO) systems using orthogonal frequency division multi-
optimize sensing technology, and as it becomes more reliable
plexing (OFDM) technology, which supports the IEEE 802.11n
and accurate for the healthcare monitoring of human beings,
protocol, provide high throughput transmission mode to serve
we can expect to see changes from the current situation.
the high data rate requirements. In such a system, disturbance of
physical objects is capable of bringing different extent variation
C. Contributions of wireless information on different subcarriers, which provided
There are a number of surveys of specific WiFi sensing conditions for the generalization of wireless sensing based on
techniques that have been published in recent years, including WiFi signals. This section, therefore, discusses some of the
human activity recognition [15]–[22], human identification [18], primary techniques used to perform WiFi sensing.
[21], localization [18], [21], vital signs [17], [18], [21], [22], 1) Received Signal Strength Indicator (RSSI): The RSSI
and imaging [21] (see Table I). All of the studies mention technique has been widely used for the localization of indi-
human activity recognition; few of them explore localization viduals. In MIMO systems, the RSSI is represented by the
and vitals estimation. The review papers discuss trends that superposition of the strength of all the received signals. Most
can be related to healthcare (human activity recognition, vitals, network devices can perform this task, including network
localization), which is introduced in Section II-C2, focusing on interface cards (NICs), as they are easily accessible. An RSSI-
the technology development, with less detail on the applications based detection system depends on the magnitude changes of
in healthcare [18]. In comparison with existing surveys within RSSI levels caused by the activity. However, due to multi-
detailed contents of various techniques and applications, the path fading and time dynamics, its performance under complex
view of our survey is distinct, which specifically focuses on conditions is significantly impacted. Early WiFi sensing systems
the analysis of WiFi applications in the healthcare field. that have been used for commercial localization are primarily
Our paper is structured to discuss the capability of WiFi dependent on RSSI without fine-grained information. Hence,
sensing in healthcare applications, including current achieve- they cannot be used to recognize complex human behavior
ments and future expectations through a thematic analysis [23].

2
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Fig. 1: Recent WiFi research used in healthcare

TABLE I: Related WiFi sensing surveys in specific track

Reference Application Range Main topic
[15] human activity recognition histogram-based techniques, deep learning methods
[16] human activity recognition comparison of human activity recognition methods
[17] respiration estimation model-based approaches, pattern-based approaches
human activity recognition,
vitals estimation, details of various WiFi sensing techniques,
[18]
localization/tracking, applications, challenges and future trends
identification
[19] human gesture recognition comparison of WiFi sensing performance
[20] human activity recognition through-wall human activity recognition methods
human activity recognition,
[21] details of various WiFi sensing techniques
imaging, vitals estimation
human activity recognition,
[22] vitals estimation, sensing applications in smart home
identification
human activity recognition, details of various WiFi sensing techniques,
This paper vitals estimation, sensing applications in healthcare,
localization technical and ethic challenges, future trends

2) Channel State Information (CSI): CSI is the channel CSI signals from all the static paths (including the signals in
property of the wireless communication link. It represents line of sight (LOS) areas and those reflected off the stationary
channel frequency response (CFR) for each subcarrier between objects). The rest of the expression is the summation of signals
transmitter and receiver, which describes the fading factor from all dynamic paths (including signals reflected from the
of the signal on every transmission path, i.e. the value of dynamic objects). Nd is the index of the dynamic path, ai (f, t)
every element in channel gain matrix H (sometimes called represents the complex attenuation factor and the initial phase of
channel matrix or channel fading matrix). In WiFi systems, the ith path; e−j2πdi (t)λ represents the phase change of ith path;
the CSI signals can be obtained from the physical layer on the di (t) and λ are the length of the ith path and the wavelength
commercial IEEE 802.11A/G/N wireless network card based of the WiFi signal, respectively. The CSI value can adapt the
on OFDM. For each subcarrier, the WiFi channel is modeled communication system to the current channel conditions and
by y = Hx + n, Where y stands for the received signal, x is guarantee high reliability and high rate communication in multi-
the transmitted signal, n is the noise component. The receiver antenna systems. With MIMO and OFDM technologies, the
computes the CSI matrix with the pre-defined signal x and the size of the CSI matrix is constructed in 3 dimensions, with N
received signal y. However, in reality, WiFi systems’ estimation transmitter antennas, M receiver antennas, and K subcarriers.
of CSI is affected by multipath fading. The CSI matrix of a The CSI packet is transmitted as N × M × K, with the packet
given subcarrier with frequency f and time t can be represented index t (see Fig. 2). The propagation performance of wireless
as [24]: signals through both the direct path and the multiple reflection
paths will show the physical space environment, including any
Nd object and the human body. Compared to RSSI values, the CSI
offers a fine-grained representation of activity. Hence recent
X
−j2π∆f t
H(f, t) = e (Hs (f ) + ai (f, t)e−j2πdi (t)λ ) (1)
i=1 device-free WiFi sensing studies favor CSI, instead of RSSI
[20].
Where e−j2π∆f t is the random phase shift due to the
hardware/software error of the WiFi system; Hs represents the

3
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Fig. 2: CSI matrices of MIMO-OFDM channels

B. Comparison with non-WiFi RF sensing represent the CFR sampling of the sub-carriers granularity
Prior to the wide use of WiFi sensing technology, consider- in the system’s frequency band, obtained from the physical
able subject-identification research has been performed with layer of the commercial IEEE 802.11n wireless network card,
traditional radar systems due to their contactless and privacy- based on OFDM technology. Based on the WiFi devices, the
preserving characteristic. For example, frequency-modulated researchers first developed an open-source CSI tool driver
continuous-wave (FMCW) radar equipment was applied in using the Intel 5300 NICs [32]. This CSI tool enables 30
[25]–[27], while in [28], the radar system was proven in its subcarriers in a 20 MHz channel bandwidth for CSI collection
accuracy through 8-meter distance monitoring of breathing and from commercial off-the-shelf (COTS) WiFi devices. This
heart rate within 5.46 - 7.25 GHz bandwidth. driver provides a quick and low-cost method to establish the
In [29], the authors use a MIMO ultra-wide-band (UWB) WiFi sensing platform. In another study, the authors in [33],
transceiver system to estimate the speed of human movement [34] have implemented their system based on the Qualcomm
with an average accuracy of 96.33%. However, in all cases, Atheros NICs offered by [35], which has 114 CSI subcarriers,
there is a high cost to establish a specific testbed. In [30], hence a higher resolution compared to the Intel 5300 CSI tool.
a system based on SDR using USRP was proposed. The In [34], the results of the comparative study have shown that
experiment simulated an FMCW system to analyze the phase the higher the number of subcarriers, the higher the sensing
change status caused by respiration. The achievements of these accuracy. Other sensing devices include the Wi-ESP which has
non-wifi sensing systems are also heavily informed by WiFi a reduced cost and is smaller in size compared to the previously
sensing due to the similarity of RF signals. However, the mentioned COTS WiFi router [36]. Besides NICs, software-
key difference between the two methods is that CSI in the defined radio (SDR) platforms are commonly used to measure
WiFi communication system is designed to recover transmitted CSI, such as the universal software radio peripheral (USRP)
information but not to explore the physical characteristic of and the wireless open-access research platform (WARP) [18],
the communication channel. For example, FMCW radar has [37], [38].
the capability to consistently and linearly adjust the frequency. 2) WiFi Sensing Applications towards Healthcare: Based
Combined with the time of flight (ToF) algorithm, FMCW on the foundation of the open-source WiFi sensing driver’s
can accurately estimate the distance information of the objects. development demonstrated in Section II-C1, researchers have
WiFi signals are only supposed to transmit within a shallow started to propose several methods and applications based
frequency bandwidth, which is limited to the devices, so it on WiFi sensing. This section provides a general overview
cannot be modulated to do the frequency sweep operation to of WiFi sensing development trends in healthcare, and more
get the range bins [31]. Nevertheless, a lot of research studies detailed technical analysis is demonstrated in Section IV. Table
found the potential of this technique and proposed various II shows some popular applications of WiFi sensing in recent
researches to compensate for the shortages and improve the years. For the convenience of demonstration, different tasks
feasibility in different tasks. are separated into two parts, human activity recognition, and
vital signs monitoring. In this case, we define the classification
and analysis of all active motion based on torso movement
C. The Evolution of WiFi Sensing for healthcare as human activity recognition. From another perspective, vital
1) Hardware platform development of WiFi sensing: For the signs are necessary to maintain regular human activity and are
past few years, research studies on CSI measurement from WiFi therefore not directly controlled by consciousness and torso
signals have been emerging for different sensing applications. movement for the vast majority of time. So, we differentiate it
In a WiFi system, CSI is essentially a data format used to from general human activity recognition.

4
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

For the human activity recognition applications in healthcare, to restore the human skeleton in visualization with only
we divide them into two types: healthcare auxiliary and WiFi signals. Combined with the state-of-the-art framework
healthcare recognition, based on the aspects of monitoring of computer vision-based human activity recognition, it has
requirement of instant and long-term feedback. In case of expected that the performance will be further improved [48],
healthcare auxiliary applications in an indoor environment, the [49].
literature covers daily activity recognition [24], [34], [39]–
[44], and other specific activity recognition such as falling III. M AIN C OMPONENTS OF C ONTACTLESS W I F I S ENSING
[45], smoking [46], sedentary behavior [47], pose estimation The development of WiFi sensing systems involves two
[48]–[51], keystroke [52] and mouth motion [53]. As for the stages, the first is applying signal processing techniques, and the
daily activity types, most papers consider: walking, running second is the algorithm design. The signal processing stage con-
(or jogging), sitting, pushing and dragging, jumping, squatting, sists of three sub-stages, i.e., denoising, signal transformation,
opening the door, and other actions that people always take in and feature extraction. The algorithm stage explains modeling-
daily life. Through these instant activity monitoring methods, based and learning-based tracks, respectively. A generalized
the alarm of dangerous accidents like falling can be transferred architecture diagram of a typical WiFi sensing system is shown
to the nearest community hospital and families to take an in Fig. 3. Firstly, raw WiFi signals are collected by the receiver
instant action to prevent delayed medical attention, especially devices, where they are denoised, transformed, and features
for elderly people [54]. At the same time, these approaches are are extracted for the data-mining of CSI signals. Secondly,
helpful for a disabled person to improve self-care capability algorithms are applied to classify/recognize/estimate the results.
through contactless interactive smart controlling methods of Each of the stages is detailed in the following subsections.
gestures recognition and pose estimation. For another range In this section, we review various kinds of technologies and
of the healthcare, recognition approaches, it mainly covers the classify them in different stages.
detection of the diseases through long-term gait monitoring, for
paraparesis detection [55] and Parkinson detection [56]. These
works train the specific model to learn the gait difference of
healthy people and patients for diseases recognition. Never-
theless, because datasets from disabled person are difficult to
obtain, relatively few works have been published in this field.
Vital signs estimation belongs to the range of healthcare
recognition applications, which is performed by monitoring
the motion of the chest and heart. Most papers analyze
the respiration rate [50], [57]–[67], some of them detect
heartbeats [50], [59], [60], [62], [64], [67], and another paper
demonstrates the biometric estimation [68]. The difference
between respiration and heartbeat estimation is demonstrated
in Section IV-B. From the perspective of potential in healthcare
applications, these systems are useful for instant monitoring
Fig. 3: WiFi Sensing System Architecture
of vitals in the non-hospital environments and helpful in the
detection of long-term chronic diseases’ such as arrhythmia,
and some respiratory diseases.
Convincing results with regards to WiFi sensing for bio- A. Signal Processing Techniques
metrics estimation is still lacking in the literature compared This stage is concerned with the processing of the collected
to radar-based systems [69], [70] which has shown good CSI signals captured during the subject’s motion. CSI data
performance. So using WiFi signals to estimate biometric is processed by different methods to obtain the nature of the
parameters can be regarded as a potential application waiting information that is required by the system. The signal process-
for further development. ing of WiFi signals constitutes three phases: Noise Reduction,
Further to the previously reported studies, it is crucial to Signal Transformation, and Feature Extraction, to feed noise-
have reliable localization and tracking systems to complement free information to the algorithms (see Section III-A1, III-A2
healthcare monitoring ones. For instance, monitoring vitals and III-A3)). Table III lists the various methodologies adopted
during sleeping cannot be performed when the person is not in and applied in the literature to process WiFi signals.
bed [71], as the position of human is essential to the decision 1) Noise Reduction: Noise components, like outliers of
making process. CSI data, always exist, which impacts the signal and causes
Meanwhile, from the timeline shown in the Table II, we a significant reduction in the recognition accuracy of the
can conclude the emerging trend is that the researchers are overall system. Denoising raw data can reduce the redundant
expecting specific WiFi sensing applications like diseases computation of invalid information and improve efficiency and
detection, biometric estimation, sedentary activity recognition, accuracy. De-noising is performed in two stages, the first is the
which have more application value in healthcare. On the other removal of outliers, and the second is performing interpolation.
hand, as the types of perceptible activities in the traditional WiFi Outlier is the data that stands out from the rest of the data
sensing method are limited, the pose estimation task is proposed set, leading to suspicion that no random deviations are resulting

5
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

TABLE II: General wifi sensing applications since from 2015 to 2020
Year of
Human Activity Recognition Vitals Signs Monitoring
Publication
gestures recognition [72], [73], human motion [74],
2015 respiration [57]
localization [75], tracking [76], daily activity [24]
localization [77], tracking [78], location and velocity [79], smoking [80], sleep vitals monitor [58],
2016
gait identification [81], mouth motion [53], gestures recognition [82] respiration and heart rate [59]
falling [45], daily activity [83], smoking [46], location and velocity [84],
2017 respiration and heart rate [60]
keystroke [52]
falling [68], dynamic velocity [85], daily activity [34], [39], [40], respiration [61],
2018 localization and tracking [86], gesture recognition [68], [87], [88], respiration and heart rate [62],
sedentary behavior [47] biometrics estimation [68]
paraparesis detection [55], localization [33], daily activity [41],
respiration [63],
2019 Parkinson detection [56], 2D pose estimation [48],
sleep vitals monitoring [64]
gesture recognition [89], [90], pedestrian flow estimation [91]
pose estimation [50], driver activity & falling [92], respiration [65], [66],
2020
3D pose estimation [49], gesture [93], [94], daily activity [42] respiration and heart rate [50], [67]

TABLE III: Signal processing techniques applied in literature for wireless sensing
Reference Noise Reduction Signal Transformation Feature Extraction Application
WiFall [45] LOF, MA N/A N/A falling detection
WiHear [53] N/A IFFT, DWT butterworth BPF mouth motion recognition
Omni-PHD [95] N/A N/A thresholding human moving detection
Somkey [46] Hampel filter, interpolation N/A thresholding smoking detection
Widar [79], [84] N/A STFT Butterworth BPF, PCA localization
Combination of Phase
PADS [85], [96] hampel filter N/A Difference and Phase human moving detection
Linear Transform
WiSee [73] Interpolation FFT band pass filter gesture recognition
Ri-2017 [97] N/A N/A signal separation by ICA daily objects moving detection
hampel filter,
TensorBeat [98] N/A thresholding vitals
PBD, SFO, CFO
CSI-Net [68] DWT, butterworth LPF N/A N/A human activity recognition
PhaseBeat [67] DWT, hampel filter FFT N/A vitals

from entirely different mechanisms. In a WiFi system, outliers two points to replace the unperceived data. Meanwhile, to keep
can be caused by hardware or software errors. Moving average the continuity of the signals, linear interpolation is applied in
(MA) is a primary method to solve the outliers, which uses many proposed systems [46], [68], [73].
statistical methods to average the CSI values in a certain period
2) Signal Transformation: The signal transformation method
and connect the average values in the time range. A Hampel
targets the analysis of CSI signals in the time-frequency domain.
filter is also used to remove the outliers, where for each sample
In the virtual environment, the wireless signal will be impacted
of the CSI datasets, the median value of the window consisting
by high and low-frequency noise. Through frequency domain
of the sample and several surrounding samples is calculated,
filtering processing, these noise signals can be effectively
and then the absolute value of the median is used to estimate
reduced. At the same time, the signal components of the
the standard deviation of the median of each sample pair. Using
frequency band required by the systems can be obtained using
the median to replace outliers is less sensitive to noise than
a band-pass filter and inverse transformation. Fast Fourier
using mean and standard deviation [18], [59]. The median filter
transform (FFT) is a standard method applied in the OFDM
has the same principle as the Hampel filter, which traverses the
systems where the CSI is a sample of FFT of channel impulse
signal without outlier detection. LOF is used to find abnormal
response (CIR). Short-time Fourier transform (STFT) frames
CSI patterns calculating the local density of the points with
and windows the original signal first, then performs FFT on
respect to k-nearest neighbors [99]. The local density of the
each frame. These characteristic assists researchers in finding
selected point will be calculated by reach-ability distance to
the dominant frequency change in the time domain, which is
neighbors and compared with other points.
efficient for real-time sensing. However, when the length of
On the other hand, interpolation processing ensures the the frame is constant, STFT takes a poor balance of signal
continuity of the signal in time and reliability of the experi- restoration in the time and frequency domains. Suppose FFT
mental data, especially when the data packets are collected at window length (for CSI signals in the time domain) gets
a higher frequency. If packets are lost during communication, extremely short, it will cause inaccurate frequency analysis with
the interpolation method would take the average of the nearest inadequate signal information. Inversely, longer window length

6
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

brings a lower resolution of signal in the time domain. Discrete

wavelet transform (DWT) is utilized to decompose signals on
different scales to improve the performance compared with
the Fourier transform. Meanwhile, DWT is available in time-
frequency analysis to judge the signal frequency changes in
the time range, the instantaneous frequency, and amplitude at
each moment.
3) Feature Extraction: Feature extraction is the process of
obtaining information from the signal, which is the basis of a
different algorithm for classification and estimation from the
CSI data. Phase difference and phase linear transform are used
to find the relationship between the changes in the phase and
human activities.
Filtering is adequate for detection of the behavior with
constant frequency like heartbeat and respiration, even for the
detection of walking, which focuses on filtered high frequency
CSI signals out to get cleaner human-related signals, such as
the respiration and heartbeat rate [67], [98]. Butterworth filter Fig. 4: Example of PCA & LPF process for CSI amplitude
is widely used because the frequency response curve in the signals
pass-band is flat without fluctuations, while it gradually drops
to zero in the stop-band.
Thresholding is used to distinguish valid signals in the value. However, due to the multi-path effect, which reflects
time range based on ToF. As shown in equation 1, the the transmitted signal, the value of ToF can be influenced.
ToF value of each path can be estimated by CSI data [78]. Power delay profile (PDP) is a common approach to get
Based on the distance of transmission lines, those signals the value of ToF through inverse fast Fourier transform
with high ToF value are reflected more times around the (IFFT), which is popular in tracking and localization [76].
environment than others, which is meaningless for systems Meanwhile, AoA makes different antennas show different phase
and can be excluded. Last but not least, signal compression observations. Multiple Signal Classification (MUSIC) performs
utilizes dimensional decrease methods that generally work in well on the AoA estimation [75], [78], which correlates phase
feature extraction, like principal component analysis (PCA) difference with the distance of multiple antennas to estimate
and independent component correlation algorithm (ICA). PCA the transmission direction.
is a statistical method, transforming a group of potentially The phase difference is required to combine with the Fresnel
correlated variables into a group of linearly uncorrelated zones (see Fig. 5). The Fresnel zone is a concentric ellipse with
variables through orthogonal transformation. This group of foci between the transmit and receiving antenna horizontally
variables after conversion is called the principal component. in the WiFi system. [100] designs the related experiments and
In WiFi sensing, PCA is mainly adopted to integrate the proves that the motion that happens in the middle of the Fresnel
signals from different subcarriers to extract main components zone is more efficient than happens on the boundary.
of variance (see Fig. 4). ICA is also a method to find the
hidden factors of non-Gaussian data, regarded as a powerful
method in blind signal analysis. From the previously familiar
sample-feature perspective, the prerequisite for using ICA is
that implicit factors of independent non-Gaussian distribution
generate the sample data.

B. WiFi Sensing Algorithms

The core methodology applied for detection or recognition
of activities lies in the algorithm, which is divided into either
modeling-based or learning-based. Some examples from the
literature are shown in the Table IV.
1) Modeling-Based Algorithm: Modeling-based approaches
apply statistical or mathematical models to extract specific Fig. 5: Geometry of the Fresnel zone
features, depending on the tasks. These studies are less
dependent on training set and have more robustness compared Doppler frequency spectrum (DFS) represents frequency
to the learning-based methods. shift influenced from the active motion, which is feasible to
ToF and angle of arrival (AoA) models have been frequently extract the velocity of subjects. CSI itself represents the channel
applied for indoor tracking and localization. When receiving frequency response, so it is convenient to do time-frequency
signals in the same physical path, the delay should be a constant analysis of the WiFi signals (see Fig. 6)). The authors of [84]

7
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

TABLE IV: Feature extraction and classification techniques applied in the literature for wireless sensing
Reference Modeling-based Learning-based Application
WiFall [45] N/A KNN, One-Class SVM falling detection
WiHear [53] MCFS DTW mouth motion recognition
Omni-PHD [95] EMD N/A human moving detection
Smokey [46] Peak Detection Autocorrelation smoking detection
Widar [79], [84] Doppler Shift,PLCR N/A localization
PADS [85], [96] N/A One-class SVM human moving detection
WiSee [73] Doppler Shift Pattern Mapping gesture recognition
CSI-Net [68] N/A Deep learning network human activity recognition
PhaseBeat [67] Peak Detection, Root-MUSIC N/A vitals estimation

defines distance measurement, which can measure the distance

between two distributions. The function of EMD and DTW
is similar in CSI based classification, which both belongs to
the range of linear regression. By contrast, DTW focus on
estimation in single dimension CSI sequence, and EMD can
be used in higher dimensions, like DFS integration shown in
Fig. 7.
Besides ML methods, many deep neural network (DNN)
frameworks have been widely applied in WiFi sensing. With
the rapid development of neural network in recent years in
Fig. 6: DFS representing human activity of leaning forward
various fields, different DNN structures have been applied
and back
in WiFi sensing, for example, convolutional neural network
(CNN), long short-term memory (LSTM) and etc. In [41],
instead of using long short-term memory (LSTM), an attention-
developed an algorithm to correlate static CSI values with based bi-directional long short-term memory (ABLSTM) neural
active multi-path gradient and utilize the Doppler frequency network is proposed to extract 2-dimensional features from
change to estimate the velocity and location of the humans. WiFi CSI data, representing human activity. In the results,
Besides, [89] adopts the body velocity profile (BVP) algorithm recognition of six different activities in public places, recorded
to apply the earth mover’s distance to integrate the multiple an accuracy of more than 97.5%. In [40] the author applies
spectrum’s characteristics to classify the gestures (see Fig. 7). the CNN classification algorithm, which records a higher
In each BVP, the velocity component is projected onto the accuracy in comparison with the SVM classifier. Similarly,
normal direction of a physical WiFi link and contributes to SignFi [87], and 1D-CNN [55] propose the use of different
the power of the corresponding radial velocity component in CNN structures for WiFi sensing. The 1D-CNN performs better
the DFS profile. Due to the path length change of the Doppler than the KNN method (average 4% higher in 1D-CNN), and
signal, WiFi equipment in a different position collects distinct SignFi improved 2% - 4% accuracy compared to the KNN-
CSI signals. This Widar3.0 considers the location of devices DTW method proposed in [82], [102]. Instead of using external
and maps DFS value into the BVP. This method reduces the neural networks (less than five layers) as alternate nonlinear
negative influence from the environment and has been tested in operators, [68] proposed a novel method to extend the size of
unknown locations where signals are collected for the training CSI input from 30 × 1 × 1 to 6 × 224 × 224 with bi-linear
set. interpolation, which provided an image-like structure for further
2) Learning-Based Algorithm: Machine learning-based clas- deep learning. It offers conditions to apply different backbone
sification algorithms such as the k-nearest nighbors (KNN) and networks like AlexNet, visual geometry group (VGG) network,
support vector machines (SVM) are widely used in detection inception cluster, etc. The deep learning network framework
and recognition tasks [101]. Multi-cluster/class feature selection proved that body perception of CSI sequences for WiFi could
(MCFS) in the WiHear system [53] sets to extract the optimal accomplish: two body characterization problems of biometrics
feature subset and find the correlation feature between different estimation (including body fat, muscle, water, and bone rate)
subsets, using a pattern matching algorithm to avoid over- and identification, two activity recognition: gesture recognition
fitting. On the other hand, with the fixed size of the dataset, and fall detection. For both traditional ML and DNN methods,
the classification process of MCFS on the testing set takes 5 the performance suffers from the distribution shift that arises
seconds, which is much lower than 3 - 5 minutes taken by the from different circumstances/locations. To avoid the repeated
SVM algorithm. The Dynamic Time Wrapping (DTW) method training of the model and fitting new areas, transfer learning
calculates the similarity between time series data by extending methods can be utilized with lower computation resources
and shortening the sequences widely used in fingerprint-based [88]. Furthermore, the metric learning approach also helps the
learning methods. Similarly, earth mover’s distance (EMD) model generalize to new environments for applications such

8
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Fig. 7: BVP methodology proposed in [89]

as gesture recognition [90]. Although all above methodologies the systems that are capable of providing smart detection
apply distinct network structures, the central task is the same, services for human beings, which can be further divided into
which is adopt DNN to match the CSI signal with the artificial two categories based on the typical tasks. Firstly, healthcare
label. auxiliary depends on the real-time detection of critical events,
Moreover, pose estimation adopts the label from camera- such as falling and irregular respiration rate, to alert the nearest
based methods, and proposes a novel DNN structure to match healthcare center or family member. Then, location and tracking
the human skeleton to WiFi CSI data. In the training stage, functions are also available in indoor healthcare auxiliary
the skeleton of a human being can be acquired from image tasks. This section reviews recent WiFi sensing healthcare
processing with cameras. Afterward, the collected WiFi data applications based on the range of supportive applications, and
is labelled and correlated with different patterns of skeleton discusses the efficiency and availability of different methods for
coordination and trained by a neural network. The authors of implementation in an indoor environment. Some applications
[103] proposed a novel network to apply a fully convolutional with technical details are shown in Table V.
network (FCN) for estimation of a single person’s pose from
the collected data and annotations. This work aims to train
the specific neural network to map the CSI variance to the A. Healthcare recognition applications
human skeletons, and get the fine-grained human skeletons from 1) Disease detection: At present, chronic disease detection
CSI signals. Furthermore, they developed another structure for in WiFi sensing is limited to activity recognition without vitals
multi-persons’ pose estimation [48]. Based on a similar theory, analysis. Due to the action difference of identity, the current
[104] proposes a image-based preprocess method to get a CSI- detection strategy is to recognize the feature of the disease
image for CNN framework to estimate the pose. However, while volunteers are doing the specific test.
in the mentioned 2D human skeleton restoration, there are The WiFreeze system [56] proposes a deep learning method
few discussions of the robustness. Due to the sensitivity of to recognize Parkinson’s disease based on the walk, sit-
CSI signals, environment has the severe impact on channel ting–standing and voluntary stopping activities of humans.
information, which means the overfitting issue is inevitable. Freezing of gait (FOG) is an explicit characteristic of Parkin-
Because 2D pixels obviously can not map all human activities, son’s disease infection. The authors apply the CSI amplitude
especially for NLOS side, with CSI variance. To improve the for continuous wavelet transform analysis and get the time-
reliablity, the study of [49] improves the BVP to 3D velocity frequency spectrograms. For the evaluation stage, they use
profile through changing the antennas’ height. These ideas the dataset of human activities that contains FOG, walking
create the condition of more applications development with slow, walking fast, sit-stand, voluntary stop. For classification,
pose estimation. they applied a revised neural network structure from VGG-
18. The result records the FOG detection accuracy of 99.7%.
IV. W I F I S ENSING APPLICATIONS FOR I N - HOME HEALTH However, the average accuracy of 5 activities is much lower,
MONITORING approximately 85.06%, which is not a good model for human
Remote healthcare monitoring systems, based on WiFi activity recognition and here are several points that needs to
sensing, perform two main operations of healthcare recog- be considered:
nition and healthcare auxiliary (described in Section II-C2). • Accounting for the phase of the CSI data and not just the
Healthcare recognition applications require the monitoring amplitude.
system to analyse the human health for activity- and vitals- • The impact of decomposing wavelets from original signals
related diseases’ detection, through the long-term monitoring when using CWT on the overall performance of a real-time
of activities and vitals. Healthcare auxiliary applications cover system.

9
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

TABLE V: WiFi sensing applications of In-home health monitoring

Reference Applications Device Type & Number Antenna Number Performance
Lower extremity paraparesis COTS WiFi device
1D-CNN [55] NTx=1, NRx=3 Average Accuracy: 99%
detection (1Rx,1Tx)
COTS WiFi device (Tx),
Breathing shortage
Wi-COVID [105] Raspberry PI 4 (Rx) NTx=1, NRx=3 Missing information
detection
(1Rx,1Tx)
COTS WiFi device
WiFreeze [56] Parkinson’s disease NTx=1, NRx=3 Average Accuracy: 99.7%
(1Rx,1Tx)
USRP B210
ESPRIT [106] Respiration rate NTx=1, NRx=1 0.15 bpm mean error
(1Rx,1Tx)
COTS WiFi device 0.6 bpm of heartbeat,
WiHealth [59] Heartbeat and Respiration rate NTx=3, NRx=3
(1Rx,1Tx) 6 bpm of respiration
0.19 bpm mean error in LOS,
COTS WiFi device
Tensorbeat [98] Respiration rate NTx=1, NRx=3 0.25 & 0.26 bpm in corridor
(1Rx,1Tx)
and through-wall scenarios.
1.19 bpm median estimation
error of heartbeat;
0.25 bpm mean error of
COTS WiFi device
PhaseBeat [67] Heartbeat and Respiration rate NTx=1, NRx=3 respiration detetion when
(1Rx,1Tx)
the sampling rate is
400Hz using directional
antennas.
In [50] Respiration rate COTS WiFi device NTx=1, NRx=3 Average Accuracy: 91.2%
Falling detection; COTS WiFi device
CSI-Net [68] NTx=3, NRx=3 Falling detection: 97.73%;
Biometrics estimation (1Rx,1Tx)
USRP N210
WiSee [73] Gestures Recognition NTx=1, NRx=1-5 Average Accuracy: 94%
(2Rx,2Tx)
Average Accuracy: 90.9%
COTS WiFi device
Widar3.0 [89] Gestures Recognition NTx=1, NRx=3 (Test Places not included
(3-6Rx,1Tx)
in the training set).
COTS WiFi device
WiFall [45] Human Activity Detection NTx=3, NRx=3 Average Accuracy: 87%
(1Rx,1Tx)
COTS WiFi device
ABLSTM [41] Human Activity Detection NTx=1, NRx=3 Average Accuracy: 99%
(1Rx,1Tx)
NTx=1, NRx=3 WiFi devices have higher
COTS WiFi device
for WiFi device; accuracy of 91% with single
WiHear [53] Mouth Detection and USRP N210
NTx=2, NRx=2 person, 74% with less than
(1Rx,1Tx)
for USRP three persons.
COTS WiFi device
CARIN [92] Driving Fatigue Detection Information missing Average F1 score: 90.9%
(1Rx,1Tx)
COTS WiFi device Average mean absolute errors
DfP [86] Localization and tracking Information missing
(3Rx,4Tx) of four environments: 0.92m
COTS WiFi device
Widar2.0 [107] Localization and tracking Information missing Average mean absolute errors: 0.75m
(1Rx,1Tx)

• The size of the training and testing datasets for DNN to the use of the 1D-CNN and the WiFreeze system cannot
avoid overfitting the model. prove the reliability of multiple persons. Secondly, the setup
details, like Barre and Mingazzini test in [55], should be more
The authors in [55] propose a 1D-CNN algorithm to detect straightforward for other researchers to repeat the experiments
lower extremity paraparesis. The dataset contains the CSI because WiFi signals can be significantly affected by the
signal of two series of specific motion detection (Barre and environment. Lastly, the discussion of application value is not
Mingazzini methods) for paraparesis detection. The 1D-CNN enough. Parkinson’s and paraparesis are chronic diseases that
system achieves an accuracy of 98% for the Barre test and should be detected in the long-term, and it has expected to do
99% for the Mingazzini test respectively. This work applies the evaluation test using a similar system on the infected groups
a relatively shallow network structure compared to the above for intelligent healthcare. Therefore, disease surveillance based
WiFreeze system. However, the dataset only contains two on WiFi sensing is a direction with development potential.
classes: typical result and abnormal result based on two test
methods. This study would improve if it considers introducing 2) Vitals detection: Vitals detection of heartbeat and respira-
more classes like gender, ages, health status to explore more tion rate belongs to passive healthcare recognition because they
based on WiFi sensing. Meanwhile, in a realistic environment, both are produced by the tiny and rhythmic vibration of the

10
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

heart and chest. This section covers the theory of respiration 150 degrees for a 5 GHz signal). For heartbeat measurement,
and heartbeat detection and related issues. it is more difficult because the motion of the heart is hugely
Respiration activity is crucial for the evaluation of sleep less than chest movements, as shown in Fig. 8. Directional
quality and the detection of respiratory diseases. One rise and antennas should be used if the system requires higher accuracy
fall of the chest is one breath, which means one inhalation of heartbeat monitoring.
and one exhalation. The regular adult breathing rate is 12 - Secondly, Fresnel zones take a significant impact on this
20 times per minute, while children and older adults have tiny movement of the heart and chest. The Fig. 9 explains how
slightly higher rates. Breathing activity has been considered the Fresnel zones influence the respiration detection. When
in several research studies, such as [57], [98]. In [98], the the people are located on the boundary of the Fresnel zone,
TensorBeat system is proposed to recognize the breathing rate the variance of phase is brutal to extract compared to the
of single and multiple individuals. The performance results of center of two boundaries. Therefore, WiFi based contactless
two-person and three-person tests are similar, accounting for vitals detection can achieve high performance but is severely
93% accuracy, with the error in both being less than 0.5 bpm. dependent on the implementation of devices.
However, the performance is reduced to approximately 62%
when number of people increased to five people, highlighting
a correlation between number of subjects and the deterioration B. Healthcare auxiliary applications
in the accuracy of breathing rate estimation. This section covers all the classification studies related to the
Heartbeat rate is also a significant index of human health. The humans’ daily activities, as discussed earlier in Section II-C2.
number of heartbeats per minute of a normal person in a calm Based on the methodologies, we divide the studies into activity
state is generally 60 to 100 beats per minute, varying among detection, pose estimation and localization. At the same time,
individuals due to age, gender, or other physiological factors. based on the movement range of human motions, the activity
In [59], [60], heart and respiratory rates are both measured detection is further separated from large-scale and small-scale,
by their proposed system. The errors are recorded for the because these tasks need to consider the specificity of actions
WiHealth system, in [59], are 0.6 bpm, for the breathing rate, when designing an experiment and the movement range of each
and 6 bpm of heart rate. PhaseBeat [67] proposes a novel task activity class in the comparison test should be kept similar. The
classification method. The authors use the resolution of the different scale will influence the experiment design as well,
feature maps to classify detected results for different tasks, shown as Fig. 10a and Fig. 10b. For instance, compared to the
which provides new ideas for future multi-task recognition gait recognition experiment, the distance and implementation
based on deep learning, and the errors decrease to 0.23 bpm of gesture detection are much more specific than the position
of respiration rate and 0.48 bpm of heartbeat rate. Besides, of gait recognition.
[67] also provides a comparison test of the omni-directional 1) Large-scale activity recognition: Large-scale recognition
antenna and directional antenna on vitals detection. In the test, methods contain falling detection, daily activity like sitting-
the error of heartbeat rate reduces to 1.19 bpm, which proves standing, gait recognition, human moving detection. Compared
the efficiency of a directional antenna. For respiration rate, with small-scale detection, large-scale recognition is generally
the error improves from 0.23 bpm to 0.25 bpm, but it is an more efficient and accessible in applications because the CSI
acceptable performance fluctuation. signals can be severely affected by the extensive range motion
of limbs and torso. For instance, in the falling detection, due to
the instantaneous velocity change of the human torso, part of
the channel transmission medium and reflection and scattering
conditions of WiFi signals have changed, resulting in the
apparent doppler effect on CSI. The WiFall system, presented
in [45], processed the amplitude of the collected CSI data to
detect falls in an indoor setting. Besides, abnormal activities
can contain health-damaged behavior like smoking, especially
in some indoor non-smoking areas. Wireless smoking behavior
testing can protect smokers’ privacy, compared to the current
camera-based detection. The WiFi wireless sensing is used
in smoking detection [46], and overcomes the blind spot in
camera-based detection systems.
Fig. 8: Heartbeats hidden in the breathing signal [67] 2) Small-scale activity recognition: Small-scale activity
detection is specific to the tiny movement of humans, including
From the comparison of results, we can conclude that the motion of hands and arms, with various kinds of studies. For
heartbeat is more difficult to sense. In WiFi frequency, 2.4 example, the authors of [94] propose a CSI-based classification
GHz and 5 GHz are standard, of which wavelengths are 60 method to classify different movements of limbs while walking.
mm and 125 mm. The maximum range of human chest motion Gesture recognition is helpful for disabled people to feasibly
while breathing is around 10 mm and 24 mm [108], which is control their daily life and contact others in an emergency [110].
far less than the wavelength. Under this condition, the phase of WiSee [73] introduces a system that sets out to recognise
reflected signals will appear a difference (from 60 degrees to nine gestures using WiFi sensing (see Fig. 11a) with up

11
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Fig. 9: Phase variety of human respiration detection in Fresnel zone boundary (Location 1) and middle region of Fresnel zone
(Location 2) of [17]

(a) Experiment setup of gesture recognition in [90] (b) Experiment setup of gait recognition in [109]

Fig. 10: Experiment setup comparison of small-scale gesture detection and large-scale gait recognition

to four people, set up in different scenarios (see Fig. 11b) for speech recognition using WiFi. The system establishes
with an average accuracy of 94%. For the tiny motion of a mapping dictionary from the pronunciation of vowels to
the hands, WiKey proposes a system to recognize stroked the mouth shapes to recognize a part of different types of
words from general keystroke behavior [52], the accuracy pronouncing and even some short words, shown in Fig. 12.
achieved was 97.5% of stroke behavior detection and 96.4% of The COTS WiFi device can capture sensitive Doppler signals in
word recognition. However, the test environment has limited CSI. Combined with the move of the head during the coughing
the transceiver direction, distance change, moving speed of process, the accuracy of cough sensing would be higher than
typing fingers and direction, the keyboard layout, and size of speaking single words. It is worth mentioning that the system
factors that affect system performance, which can be difficult adopts directional antennas for implementation.
to replicate the work in a real-life scenario. To improve the
robustness, [89] is proposed to correlate the CSI data of 3) Fine-grained human pose estimation: In daily life, human
more than three receivers in different directions and positions. body language is extremely various that it is difficult to
The result shows average accuracy above 85% for all testing categorize it with a few fixed variables. Therefore, a system
locations. Nevertheless, the good performance of the system that uses limited data sets for motion classification is difficult
encourages researchers to design more applications based on to deploy effectively in a non-experimental environment. Pose
WiFi signals. estimation targets to recover the human skeleton for people
to recognize the human actions through the vision, which
Mouth movement always represents speaking, eating, and is clear and intuitive. Compared to the classification studies
coughing. In the healthcare monitoring system, cough is mentioned above, this fine-grained pose estimation is more
a significant symptom in patients infected with respiratory competitive and humanized and capable of recovering both
diseases, including COVID-19 [111], [112]. Wireless-based large-scale and small-scale activities’ pose. A fine-grained
activity recognition can be regarded as a suitable method for [48] system was firstly implemented in WiFi range with the
cough detection. Recently, WiFi sensing has been applied skeleton on the 2D images for users, shown in Fig. 13. This
in many other health-related applications, other than cough system tests processed human skeletons and achieves 0.66 of
detection. On this side, [53] presents the WiHear system mean intersection over union (mIoU) with 1 - 5 persons. [49]

12
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

(a) Gestures diagrams

(b) Testbed diagrams

Fig. 11: Nine gestures detected in WiSee System in 11a and scenarios of testbed in Whole-home range in 11b of [73]

proposes a 3D skeleton restoration method using WiFi signals, 4) Human Localization: Further to the previously high-
where the authors used the VICON motion camera system as lighted studies, WiFi sensing also shows potential in indoor
the ground truth 3D skeleton of the human body, shown in Fig. localization and tracking. At present, traditional WiFi localiza-
14. In this system, the skeleton represents the physical distance tion adopts RSSI signals from mobile devices, which depends
instead of the pixels, which further proves that the WiFi signal multiple access points (AP) with different media access control
can perform fine-grained detection. The restored skeleton can address value for measurement (at least three devices). CSI
assist users in judging the performance of the sensing system based WiFi localization can reduce the required number of
and is available for further activity recognition. Also, due to AP, and be independent on portable devices. In [113], the
visible skeleton restoration, these works have the exemplary authors propose a method with location fingerprint to locate
significance regarding the increase of the human activity types people with RSSI fingerprints using the WiFi COTS devices
without re-training the whole system. However, human skeleton and mobile phones. In the setup shown in Fig. 15, there are
restoration is limited to the overfitting. The training set of 3D 24 APs in total equipped in a plain floor of 1610 square meter.
restoration works only contains the human skeleton in constant DeMan [74] regards human breathing as an inherent indicator
location, which means most predicted skeletons have limited of the human state and judges the existence of a stationary
variance. person by detecting specific signal patterns caused by tiny chest

13
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Fig. 15: RSSI based localization implementation of [113]

movements. Another work proposed a passive localization

indoor system [77] using CSI with an improved Dynamic-
MUSIC algorithm. The system gets the AoA value from moving
targets to establish the fingerprints of the location (see Fig.
16) and records a median error of approximately 0.6 m, with
Fig. 12: Mouth motion detection in WiHear [53] two pair of devices, which is sufficient for localizing children
and disabled persons. In [86], a probabilistic fingerprint-based
method is proposed and tested in four environments. Their
result demonstrates the system of the Kalman filter for tracking
objects at a fixed walking speed, which is available to analyze
the movement of monitored people with pose estimation for
early symptom detection, with 3 transmitters and four receivers.
More specifically, [107] achieves the indoor tracking with
one pair of WiFi devices, in general office and corridor, with
the average error of 1 meter. In conclusion, contactless WiFi
sensing approach can be applied with fewer number of devices
compared to RSSI based methods, without portable devices.
Fig. 13: 2D restoration of human pose [48] However, RSSI-based method is more capable to achieve the
larger range tracking in cross-rooms environment.

V. C HALLENGES AND F UTURE D IRECTIONS

Due to the popularity and consistent upgrade of WiFi devices,
WiFi sensing is the technology with considerable potential.
However, the popularisation of technology in healthcare will
inevitably encounter not only technical limitations as appli-
cations introduced but also ethical problems in Section IV.
Specific challenges and future directions are presented and
discussed in this section.

A. Challenges of WiFi sensing in healthcare

Section IV has introduced various types of WiFi sensing
systems for healthcare that achieved high accuracy in a specific
task. However, most of the work is challenging to implement
at home simply because the environment is distinct, which
means the performance is limited with regards to different
scenarios (see the comparison regarding six different factors
in Table VI,VII,VIII,IX,X). This section discusses these issues
and gives the corresponding technical-level solutions.
1) Robustness: Robustness is essential for the future devel-
opment of WiFi healthcare sensing, resulting in whether the
experiment’s repeatability and validity can be guaranteed in
different environments. Although most learning-based methods
Fig. 14: 3D restoration of human pose [49] have good experimental results, due to the method itself
on the specific data, the performance of such methods in

14
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

(a) Meeting room (b) Office room (c) Lobby

Fig. 16: CSI based localization system environments of [77]

TABLE VI: Sampling frequency versus WiFi sensing performance

Smokey [46] widar2.0 [107] DfP [86] PhaseBeat [67]
no significant change on
no significant change
Increase 20% accuracy breathing detection,
Sampling frequency drop 80% (100Hz - 10Hz) (time consumption increases
(1kHz-10kHz) heart rate error decreases
from 0.7s to 1.5s)
from 3bpm to 1.1bpm

TABLE VII: Performing distance versus WiFi sensing performance

Human
MaTrack
WiFall [45] Smokey [46] recognition WiGest [72] Widar3.0 [89] RespiRadio [114] PhaseBeat [67]
[77]
[81]
the error increases the error increases
Distance 13% maximum
drop 16% drop 33% drop 45% drop 20% drop 20% from 0.1bpm to from 0.27bpm to
or difference
(1.5m - 3m) (0.5m - 3m) (2m - 16m) (1m - 19m) (1m - 9m) 1.2bpm 0.54bpm
location of 5 locations
(Tx-Rx: 2m - 10m) (Tx-Rx: 2m - 7m)

TABLE VIII: Performing environments versus WiFi sensing performance

WiFall [45] Smokey [46] SignFi [87] Widar3.0 [89] DfP [86]
9% maximum 19.2% maximum 3% maximum 4.5% maximum 0.2m maximum
Environments difference of difference of difference of difference of error of 4
3 environments 3 environments 3 environments 3 environments environments

unknown scenarios cannot be promised. In the comparison 2) Complicated implementation of setup: Meanwhile, it is
tables, most challenges are related to achieving generalization also a challenge to set up the WiFi devices while using. In [89],
of sensing methods under different conditions: distance or the study involves a robust system for gesture recognition using
location, environments, identity difference, and orientation of 3 - 6 receivers with the predetermined location. Objectively, it’s
action. These factors are discussed in the studies and show the not possible for each consumer to set the devices in the same
limitation behind the overall performance. position due to the limited physical spaces or other issues.
In addition, sampling frequency’s setting has a significant
Therefore, sensitivity to changes in the physical environment impact on performance as shown in Table VI, which needs
is a double-edged sword, which supports the high performance experiment validation to achieve a trade-off between low power
of detecting specific activity but allows much noise from consumption and high accuracy performance. These problems
surroundings to disturb the process. For example, for the have hindered the popularisation of the WiFi sensing technique.
methods using DNN and ML for activity recognition, the
size of the experimental setup is kept in a limited variance of 3) Multiple Subjects Sensing: The performance of different
the noise. For the complex real-world environment, the barriers WiFi sensing systems becomes worse as the number of subjects
around the circumstance can take serious adverse effects on involved in the experiment increases, as shown in Table X.
the accuracy. For example, estimation of human respiration Majority of works do not mention the multiple human scenarios
rate needs to filter out the noise component of other activities. due to the low angle resolution of WiFi. Although the AoA
However, even if unconscious human activity is filtered out, technique based on MIMO is accurate enough to distinguish
conscious rhythmic activity can easily interfere with detection the human and count the number of people, it is not enough
results and other objects’ activities. to distinguish the signal from a different person and get a

15
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

TABLE IX: Performing environments versus WiFi sensing performance

Gait recognition [115] WiKey [52] WiFinger [82] SignFi [87] Widar3.0 [89] widar2.0 [107] PhaseBeat [67]
15% maximum 14% maximum 30% maximum 13% maximum 0.2bpm maximum
Identity average accuracy no significant
difference of difference of difference of difference of difference of
difference drop 24% (2-11) change
10 subjects 10 subjects 5 subjects 8 subjects 4 directions

TABLE X: Orientation and Multiple human versus WiFi sensing performance

WiFall [45] Widar3.0 [89] Smokey [46] PhaseBeat [67]
error increases from
drop 30% 11% maximum difference drop 25%
Orientations Multiple human senarios 0.25 bpm to 0.7 bpm
(10 - 110 degree) of 5 directions (1 - 4)
(2 - 4)

specific result. To improve the resolution of sensing, efficient which has a completely unique setup and test standard for
multiple subjects recognition method must be developed and different methodologies to assess their performance in the
improved for more practical and real-world setups. Meanwhile, indoor environment. For example, the framework should contain
it is challenging to track the person’s identity with specific the available sampling frequency, type of antennas and NICs,
respiration for in-home healthcare analysis in the multiple human location, distance between transmitter and receiver,
subjects environment. number of subjects and etc. Based on the framework, extra
4) Performance of WiFi Networking: CSI is collected using modules for mobile edge computing can be incorporated in the
NICs, which are primarily used for networking, with stable WiFi sensing systems [93] to accelerate the processing speed
transmitted frequency. Current drivers of WiFi device is able and decrease the influence on WiFi networking system.
to increase the frequency to meet the requirement of higher 2) Spatial Sensing for wide applications: WiFi sensing
frequency of CSI packets collection that most of the systems applications can sense ambient information regardless of
rely on, which will produce empty packets for networking. In direction due to the omni-directional antennas of WiFi devices.
the case where the empty package takes up more, they will However, this characteristic is not helpful in some cases. If
interfere with typical networking tasks. Hence, it is challenging the WiFi sensing system can monitor the vitals or activities of
to operate a WiFi device to complete sensing and networking individuals, for example, only detect the vital signs of people
tasks simultaneously. on the fixed bed in the hospital, instead of detecting caregivers.
5) Privacy and Security: Privacy and security are being It will be more efficient to analyze the data without massive
threatened by WiFi passive sensing technology. In state-of-the- ambient noise from other directions. Nevertheless, due to CSI
art methodologies, WiFi sensing can be used in the NLOS data collected from WiFi devices that have low resolution and
range for human and object behavior. A well-trained system can high sensitivity, it is challenging to separate signals from a
be used to recognize the gesture and keystrokes of the human specific space. Beamforming technology with an intelligent
being. Suppose WiFi sensing is used to steal other people’s reflecting surface (IRS) [116] and directional antennas [67]
private information due to the portability and generalization are considered methods to overcome this disadvantage and
of WiFi equipment. In that case, it is difficult for people to improve the performance.
recognize privacy-invasion behavior from those who collect With spatial sensing, WiFi sensing of human activities is
that private information. Fortunately, thanks to the limitations not limited to residential environments only. For instance, they
of NIC manufacturers on channel information decoding, only a can be of great use in vehicles to detect drivers’ tiredness
limited number of devices can collect CSI data through open- levels, which is significant for the safety of the driver and
source drivers. However, with the miniaturization of NICs and the passengers. The authors of [92] have successfully set
the maturity of sensing technology, this issue will be much up the WiFi testbed in a vehicle, and eight human activities
more concerning in the future. from the driver and the passengers were accounted for in the
CARIN system, including pushing, pulling, and swiping. The
B. Future Directions of WiFi Sensing in Healthcare experiment results have shown an average accuracy of 90.0%
Although WiFi sensing has significant challenges for gener- for more than 3000 real-world activity traces. Also, the vitals
alization, there is still great potential for healthcare applications detection in-cabin can be more efficient than the detection
in the real world. To overcome the issues discussed above, the indoor because the position of people in the traffic tools is
technique is expected to improve in four directions. constant in most cases. It will not take effort to change sensing
1) A unified framework of WiFi sensing: One of the most area with spatial sensing like directional antennas are others.
challenging problems of WiFi sensing robustness is various 3) Joint sensing in Home for Healthcare: Implementation
experimental setup and devices. Although theoretically, all in real-world environments of WiFi sensing is one of the most
WiFi sensing methods should get the same performance as any challenging tasks. Although many model-based algorithms
type of WiFi device. In practice, even the sampling rate of described in Section III-B2 provide the approaches of human
WiFi can significantly influence the performance (see Table action estimation without training, especially in vitals detection
VI). It is necessary to follow a framework from hardware and tracking. For activity recognition, most studies still adopt
to software of WiFi sensing system for future generalization, DNN based classification method to get the result with better

16
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

performance due to the influence of unexpected human actions. [3] L. P. Malasinghe, N. Ramzan, and K. Dahal, “Remote patient
The trends of joint sensing can be separated into two aspects, monitoring: a comprehensive study,” Journal of Ambient Intelligence
and Humanized Computing, vol. 10, no. 1, pp. 57–76, 2019. [Online].
the first one is joint training and second one is joint sensing Available: http://dx.doi.org/10.1007/s12652-017-0598-x
[117]. Joint training will combine the WiFi signals and different [4] Q. Huang, H. Chen, and Q. Zhang, “Joint Design of Sensing and
types of data from other sources like skeleton key-points from Communication Systems for Smart Homes,” IEEE Network, vol. 34,
no. 6, pp. 191–197, 2020.
images as ground truth datasets for the training step of the [5] S. Majumder, T. Mondal, and M. J. Deen, “Wearable sensors for remote
estimation system with WiFi signals. Hence, WiFi signals can health monitoring,” Sensors (Switzerland), vol. 17, no. 1, 2017.
be used in other tasks with higher resolution and more functions [6] D. Xiong, D. Zhang, X. Zhao, and Y. Zhao, “Deep learning for emg-
based human-machine interaction: A review,” IEEE/CAA Journal of
in human support without label limitation. On the other hand, Automatica Sinica, vol. 8, no. 3, pp. 512–533, 2021.
DNN is available in large size of training and mining the [7] M. Andriluka, L. Pishchulin, P. Gehler, and B. Schiele, “2d human
feature from data for multiple human sensing [48]. pose estimation: New benchmark and state of the art analysis,” in 2014
IEEE Conference on Computer Vision and Pattern Recognition, 2014,
Joint sensing needs to establish a whole framework of general pp. 3686–3693.
indoor environments for the users. Due to the restriction of [8] M. B. Khan, Z. Zhang, L. Li, W. Zhao, M. A. M. Al Hababi, X. Yang,
performance versus WiFi devices’ setting, as demonstrated in and Q. H. Abbasi, “A systematic review of non-contact sensing for
Section V-A, CSI data cannot be used to perform multiple developing a platform to contain covid-19,” Micromachines, vol. 11,
no. 10, pp. 1–23, 2020.
tasks within one type of implementation in complex real-world [9] M. Alizadeh, G. Shaker, and S. Safavi-Naeini, “Remote heart rate
scenarios. Extensive data analysis has been the trend for the sensing with mm-wave radar,” in 2018 18th International Symposium
future healthcare [118]. CSI signal can be integrated with data on Antenna Technology and Applied Electromagnetics (ANTEM), 2018,
pp. 1–2.
from other sensors to monitor the health status of human beings [10] F. Khan, A. Ghaffar, N. Khan, and S. H. Cho, “An overview of signal
using artificial intelligence techniques [4], [68]. Cooperative processing techniques for remote health monitoring using impulse radio
sensing with different sensors will improve the precision with UWB transceiver,” Sensors (Switzerland), vol. 20, no. 9, 2020.
[11] W. Taylor, S. A. Shah, K. Dashtipour, A. Zahid, Q. H. Abbasi, and
a larger size of precise data. Different kinds of radar sensors M. A. Imran, “An intelligent non-invasive real-time human activity
(millimeter-wave radar, LIDAR), and other sensors like infrared recognition system for next-generation healthcare,” Sensors, vol. 20,
sensors, cameras, microphones, are able to provide data for no. 9, p. 2653, 2020.
[12] J. Qi, P. Yang, L. Newcombe, X. Peng, Y. Yang, and Z. Zhao, “An
users to achieve specific tasks [50], [119], [120]. Joint sensing overview of data fusion techniques for internet of things enabled physical
around the household environment is potential to improve the activity recognition and measure,” Information Fusion, vol. 55, pp. 269–
humanization of the future healthcare monitoring system in 280, 2020.
[13] A. Rashed, A. Ibrahim, A. Adel, B. Mourad, A. Hatem, M. Magdy,
the whole-home range. N. Elgaml, and A. Khattab, “Integrated iot medical platform for remote
healthcare and assisted living,” in 2017 Japan-Africa Conference on
VI. C ONCLUSIONS Electronics, Communications and Computers (JAC-ECC), 2017, pp.
160–163.
Intelligent sensing based on commercial WiFi devices is [14] Y. Ge, S. Ansari, A. Abdulghani, M. A. Imran, and Q. H. Abbasi,
an emerging trend for next-generation healthcare monitoring “Intelligent instruction-based iot framework for smart home applications
using speech recognition,” in 2020 IEEE International Conference on
with great potential of development, due to the low-cost and Smart Internet of Things (SmartIoT), 2020, pp. 197–204.
popularization of WiFi devices. Meanwhile, WiFi sensing is [15] “A Survey on Behavior Recognition Using WiFi Channel State
capable to provide high accuracy contactless monitoring service Information,” IEEE Communications Magazine, vol. 55, no. 10, pp.
98–104, 2017.
for the users, unlike other wearable sensors and systems that [16] Z. Wang, B. Guo, Z. Yu, and X. Zhou, “Wi-Fi CSI-Based Behavior
require intimate contact. This paper presented an extensive Recognition: From Signals and Actions to Activities,” IEEE Communi-
review of the recent techniques, open research challenges, cations Magazine, vol. 56, no. 5, pp. 109–115, 2018.
[17] D. Wu, D. Zhang, C. Xu, H. Wang, and X. Li, “Device-Free WiFi
and development trends of non-contact WiFi sensing related Human Sensing: From Pattern-Based to Model-Based Approaches,”
to healthcare. Different classes of healthcare applications are IEEE Communications Magazine, vol. 55, no. 10, pp. 91–97, 2017.
demonstrated, including large-scale and small-scale activity [18] Y. Ma, G. Zhou, and S. Wang, “WiFi sensing with channel state
information: A survey,” ACM Computing Surveys, vol. 52, no. 3, 2019.
recognition like falling and gestures, vital signs’ detection, [19] “Device free human gesture recognition using Wi-Fi CSI: A
and supportive localization. These clinical applications are survey,” Engineering Applications of Artificial Intelligence, vol. 87,
supportive to reduce the future pressure from the rapidly rising no. January 2019, p. 103281, 2020. [Online]. Available: https:
//doi.org/10.1016/j.engappai.2019.103281
aging population presents numerous societal and economic
[20] Z. Wang, K. Jiang, Y. Hou, Z. Huang, W. Dou, C. Zhang, and Y. Guo,
challenges. “A Survey on CSI-Based Human Behavior Recognition in Through-The-
Wall Scenario,” IEEE Access, vol. 7, pp. 78 772–78 793, 2019.
[21] Y. He, Y. Chen, Y. Hu, and B. Zeng, “WiFi Vision: Sensing, Recognition,
ACKNOWLEDGMENT and Detection with Commodity MIMO-OFDM WiFi,” IEEE Internet
This work is supported in parts by EPSRC grant no. of Things Journal, vol. 7, no. 9, pp. 8296–8317, 2020.
[22] H. Jiang, C. Cai, X. Ma, Y. Yang, and J. Liu, “Smart Home Based
EP/T021020/1, EP/T021063/1, and EP/W003228/1. on WiFi Sensing: A Survey,” IEEE Access, vol. 6, pp. 13 317–13 325,
2018.
R EFERENCES [23] Y. Wang, M. Li, and M. Li, “The statistical analysis of ieee 802.11
wireless local area network–based received signal strength indicator in
[1] P.-Y. Hsu, P.-H. Hsu, T.-H. Lee, and H.-L. Liu, “Heart rate and indoor location sensing systems,” International Journal of Distributed
respiratory rate monitoring using seismocardiography,” in 2021 43rd Sensor Networks, vol. 13, no. 12, p. 1550147717747858, 2017.
Annual International Conference of the IEEE Engineering in Medicine [24] W. Wang, A. X. Liu, M. Shahzad, K. Ling, and S. Lu, “Understanding
Biology Society (EMBC), 2021, pp. 6876–6879. and modeling of wifi signal based human activity recognition,” in
[2] S. Patel, H. Park, P. Bonato, L. Chan, and M. Rodgers, “A review of Proceedings of the 21st Annual International Conference on Mobile
wearable sensors and systems with application in rehabilitation,” 2012. Computing and Networking, ser. MobiCom ’15. New York, NY,

17
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

USA: Association for Computing Machinery, 2015, p. 65–76. [Online]. [45] Y. Wang, K. Wu, and L. M. Ni, “WiFall: Device-Free Fall Detection by
Available: https://doi.org/10.1145/2789168.2790093 Wireless Networks,” IEEE Transactions on Mobile Computing, vol. 16,
[25] S. A. Shah, A. Tahir, J. Ahmad, A. Zahid, H. Pervaiz, S. Y. Shah, no. 2, pp. 581–594, 2017.
A. M. A. Ashleibta, A. Hasanali, S. Khattak, and Q. H. Abbasi, “Sensor [46] X. Zheng, J. Wang, L. Shangguan, Z. Zhou, and Y. Liu, “Design and
fusion for identification of freezing of gait episodes using wi-fi and radar implementation of a csi-based ubiquitous smoking detection system,”
imaging,” IEEE Sensors Journal, vol. 20, no. 23, pp. 14 410–14 422, IEEE/ACM Transactions on Networking, vol. 25, no. 6, pp. 3781–3793,
2020. 2017.
[26] Y. Song, W. Taylor, Y. Ge, K. Dashtipour, M. A. Imran, and Q. H. [47] J. Yang, H. Zou, H. Jiang, and L. Xie, “CareFi: Sedentary Behav-
Abbasi, “Design and implementation of a contactless ai-enabled human ior Monitoring System via Commodity WiFi Infrastructures,” IEEE
motion detection system for next-generation healthcare,” in 2021 IEEE Transactions on Vehicular Technology, vol. 67, no. 8, pp. 7620–7629,
International Conference on Smart Internet of Things (SmartIoT), 2021, 2018.
pp. 112–119. [48] F. Wang, S. Zhou, S. Panev, J. Han, and D. Huang, “Person-in-wifi: Fine-
[27] A. Shrestha, H. Li, J. Le Kernec, and F. Fioranelli, “Continuous human grained person perception using wifi,” in Proceedings of the IEEE/CVF
activity classification from fmcw radar with bi-lstm networks,” IEEE International Conference on Computer Vision (ICCV), October 2019.
Sensors Journal, vol. 20, no. 22, pp. 13 607–13 619, 2020. [49] W. Jiang, H. Xue, C. Miao, S. Wang, S. Lin, C. Tian, S. Murali,
[28] F. Adib, H. Mao, Z. Kabelac, D. Katabi, and R. C., “Smart homes H. Hu, Z. Sun, and L. Su, “Towards 3D human pose construction using
that monitor breathing and heart rate,” in ACM Conference on Human wifi,” Proceedings of the Annual International Conference on Mobile
Factors in Computing Systems, 2015, pp. 837–846. Computing and Networking, MOBICOM, pp. 295–308, 2020.
[29] Y. Chen, H. Deng, D. Zhang, and Y. Hu, “Speednet: Indoor speed [50] L. Guo, Z. Lu, S. Zhou, X. Wen, and Z. He, “When healthcare meets
estimation with radio signals,” IEEE Internet of Things Journal, vol. 8, off-the-shelf wifi: A non-wearable and low-Costs approach for in-home
no. 4, pp. 2762–2774, 2021. monitoring,” arXiv, 2020.
[30] A. M. Ashleibta, A. Zahid, S. A. Shah, Q. H. Abbasi, and M. A. Imran, [51] S. Zhou, L. Guo, Z. Lu, X. Wen, W. Zheng, and Y. Wang, “Subject-
“Flexible and scalable software defined radio based testbed for large independent human pose image construction with commodity wi-fi,” in
scale body movement,” Electronics, vol. 9, no. 9, p. 1354, 2020. ICC 2021 - IEEE International Conference on Communications, 2021,
[31] A. Meta, P. Hoogeboom, and L. P. Ligthart, “Signal processing for pp. 1–6.
fmcw sar,” IEEE Transactions on Geoscience and Remote Sensing, [52] K. Ali, A. X. Liu, W. Wang, and M. Shahzad, “Recognizing Keystrokes
vol. 45, no. 11, pp. 3519–3532, 2007. Using WiFi Devices,” IEEE Journal on Selected Areas in Communica-
tions, vol. 35, no. 5, pp. 1175–1190, 2017.
[32] D. Halperirr, W. Hu, A. Sheth, and D. Wetherall, “Tool release: gathering
802.11n traces with channel state information,” Acm Sigcomm Computer [53] G. Wang, Y. Zou, Z. Zhou, K. Wu, and L. M. Ni, “We Can Hear You
Communication Review, vol. 41, no. 1, pp. 53–53, 2011. with Wi-Fi!” IEEE Transactions on Mobile Computing, vol. 15, no. 11,
pp. 2907–2920, 2016.
[33] Y. Xie, Z. Li, and M. Li, “Precise power delay profiling with commodity
[54] M. M. Hassan, A. Gumaei, G. Aloi, G. Fortino, and M. Zhou, “A
wi-fi,” IEEE Transactions on Mobile Computing, vol. 18, no. 6, pp.
smartphone-enabled fall detection framework for elderly people in
1342–1355, 2019.
connected home healthcare,” IEEE Network, vol. 33, no. 6, pp. 58–63,
[34] J. Yang, H. Zou, H. Jiang, and L. Xie, “Device-Free Occupant Activity
2019.
Sensing Using WiFi-Enabled IoT Devices for Smart Homes,” IEEE
[55] L. Guan, F. Hu, F. Al-Turjman, M. B. Khan, and X. Yang, “A Non-
Internet of Things Journal, vol. 5, no. 5, pp. 3991–4002, 2018.
Contact Paraparesis Detection Technique Based on 1D-CNN,” IEEE
[35] Y. Xie, Z. Li, and M. Li, “Precise power delay profiling with Access, vol. 7, pp. 182 280–182 288, 2019.
commodity wifi,” in Proceedings of the 21st Annual International
[56] A. Tahir, J. Ahmad, S. A. Shah, G. Morison, D. A. Skelton, H. Larijani,
Conference on Mobile Computing and Networking, ser. MobiCom
Q. H. Abbasi, M. A. Imran, and R. M. Gibson, “WiFreeze : Multires-
’15. New York, NY, USA: ACM, 2015, p. 53–64. [Online]. Available:
olution Scalograms for Freezing of Gait Detection in Parkinson ’ s
http://doi.acm.org/10.1145/2789168.2790124
Leveraging 5G Spectrum with Deep Learning,” pp. 1–18, 2019.
[36] M. Atif, S. Muralidharan, B. Yoo, and H. Ko, “Wi-ESP — A tool for [57] H. Abdelnasser, K. A. Harras, and M. Youssef, “Ubibreathe: A ubiqui-
CSI-based Device-Free Wi-Fi Sensing ( DFWS ),” vol. 7, no. May, pp. tous non-invasive wifi-based breathing estimator,” in ACM International
644–656, 2020. Symposium on Mobile Ad Hoc Networking and Computing, 2015, pp.
[37] M. Rehman, R. A. Shah, M. B. Khan, N. A. A. Ali, A. A. Alotaibi, 277—-286.
T. Althobaiti, N. Ramzan, S. A. Shah, X. Yang, A. Alomainy, M. A. [58] P. Nguyen, X. Zhang, A. Halbower, and T. Vu, “Continuous and
Imran, and Q. H. Abbasi, “Contactless small-scale movement monitoring fine-grained breathing volume monitoring from afar using wireless
system using software defined radio for early diagnosis of covid-19,” signals,” in IEEE INFOCOM 2016 - The 35th Annual IEEE International
IEEE Sensors Journal, vol. 21, no. 15, pp. 17 180–17 188, 2021. Conference on Computer Communications, 2016, pp. 1–9.
[38] S. A. K. Tanoli, M. Rehman, M. B. Khan, I. Jadoon, F. Ali Khan, [59] J. Shang and J. Wu, “Fine-grained vital signs estimation using
F. Nawaz, S. A. Shah, X. Yang, and A. A. Nasir, “An experimental commercial wi-fi devices,” in Proceedings of the Eighth Wireless of
channel capacity analysis of cooperative networks using universal the Students, by the Students, and for the Students Workshop, ser. S3.
software radio peripheral (usrp),” Sustainability, vol. 10, no. 6, p. 1983, New York, NY, USA: Association for Computing Machinery, 2016, p.
2018. 30–32. [Online]. Available: https://doi.org/10.1145/2987354.2987360
[39] B. Tan, Q. Chen, K. Chetty, K. Woodbridge, W. Li, and R. Piechocki, [60] X. Wang, C. Yang, and S. Mao, “Phasebeat: Exploiting csi phase data for
“Exploiting wifi channel state information for residential healthcare vital sign monitoring with commodity wifi devices,” in 2017 IEEE 37th
informatics,” IEEE Communications Magazine, vol. 56, no. 5, pp. 130– International Conference on Distributed Computing Systems (ICDCS),
137, 2018. 2017, pp. 1230–1239.
[40] T. Z. Chowdhury, “Using Wi-Fi channel state information (CSI) [61] C. Chen, Y. Han, Y. Chen, H.-Q. Lai, F. Zhang, B. Wang, and K. J. R.
for human activity recognition and fall detection,” pp. 24–50, Liu, “Tr-breath: Time-reversal breathing rate estimation and detection,”
2018. [Online]. Available: https://open.library.ubc.ca/cIRcle/collections/ IEEE Transactions on Biomedical Engineering, vol. 65, no. 3, pp.
ubctheses/24/items/1.0365967 489–501, 2018.
[41] Z. Chen, L. Zhang, C. Jiang, Z. Cao, and W. Cui, “WiFi CSI Based [62] S. Lee, Y.-D. Park, Y.-J. Suh, and S. Jeon, “Design and implementation
Passive Human Activity Recognition Using Attention Based BLSTM,” of monitoring system for breathing and heart rate pattern using
IEEE Transactions on Mobile Computing, vol. 18, no. 11, pp. 2714– wifi signals,” in 2018 15th IEEE Annual Consumer Communications
2724, 2019. Networking Conference (CCNC), 2018, pp. 1–7.
[42] S. Zhou, W. Zhang, D. Peng, Y. Liu, X. Liao, and H. Jiang, “Adversarial [63] Y. Zeng, D. Wu, J. Xiong, E. Yi, R. Gao, and D. Zhang, “Farsense:
wifi sensing for privacy preservation of human behaviors,” IEEE Pushing the range limit of wifi-based respiration sensing with csi ratio
Communications Letters, vol. 24, no. 2, pp. 259–263, 2020. of two antennas,” no. 121, 2019.
[43] Q. Chen, Y. Liu, B. Tan, K. Woodbridge, and K. Chetty, “Respiration [64] Y. Gu, X. Zhang, Z. Liu, and F. Ren, “Wifi-based real-time breath-
and Activity Detection Based on Passive Radio Sensing in Home ing and heart rate monitoring during sleep,” in 2019 IEEE Global
Environments,” IEEE Access, vol. 8, pp. 12 426–12 437, 2020. Communications Conference (GLOBECOM), 2019, pp. 1–6.
[44] Y. Ge, S. Li, M. Shentu, A. Taha, S. Zhu, J. Cooper, M. Imran, and [65] M. Ibrahim and K. N. Brown, “Vehicle in-cabin contactless wifi human
Q. Abbasi, “A doppler-based human activity recognition system using sensing,” in 2021 18th Annual IEEE International Conference on
wifi signals,” in 2021 IEEE Sensors, 2021, pp. 1–4. Sensing, Communication, and Networking (SECON), 2021, pp. 1–2.

18
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

[66] W. Li, R. J. Piechocki, K. Woodbridge, C. Tang, and K. Chetty, “Passive [86] S. Shi, S. Sigg, L. Chen, and Y. Ji, “Accurate Location Tracking from
wifi radar for human sensing using a stand-alone access point,” IEEE CSI-Based Passive Device-Free Probabilistic Fingerprinting,” IEEE
Transactions on Geoscience and Remote Sensing, vol. 59, no. 3, pp. Transactions on Vehicular Technology, vol. 67, no. 6, pp. 5217–5230,
1986–1998, 2021. 2018.
[67] X. Wang, C. Yang, and S. Mao, “On csi-based vital sign monitoring [87] Y. Ma, G. Zhou, S. Wang, H. Zhao, and W. Jung, “SignFi: Sign
using commodity wifi,” ACM Trans. Comput. Healthcare, vol. 1, no. 3, Language Recognition Using WiFi,” Proceedings of the ACM on
May 2020. [Online]. Available: https://doi.org/10.1145/3377165 Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 2,
[68] F. Wang, J. Han, S. Zhang, X. He, and D. Huang, “CSI-Net: Unified no. 1, pp. 1–21, 2018.
Human Body Characterization and Pose Recognition,” arXiv1810.03064, [88] H. Zou, J. Yang, Y. Zhou, L. Xie, and C. J. Spanos, “Robust WiFi-
no. 2, 2018. enabled device-free gesture recognition via unsupervised adversarial
[69] D. Rissacher and D. Galy, “Cardiac radar for biometric identification domain adaptation,” Proceedings - International Conference on Com-
using nearest neighbour of continuous wavelet transform peaks,” in IEEE puter Communications and Networks, ICCCN, vol. 2018-July, 2018.
International Conference on Identity, Security and Behavior Analysis [89] Y. Zheng, Y. Zhang, K. Qian, G. Zhang, Y. Liu, C. Wu, and
(ISBA 2015), 2015, pp. 1–6. Z. Yang, “Zero-effort cross-domain gesture recognition with wi-fi,” in
[70] K. Diederichs, A. Qiu, and G. Shaker, “Wireless biometric individual Proceedings of the 17th Annual International Conference on Mobile
identification utilizing millimeter waves,” IEEE Sensors Letters, vol. 1, Systems, Applications, and Services, ser. MobiSys ’19. New York,
no. 1, pp. 1–4, 2017. NY, USA: Association for Computing Machinery, 2019, p. 313–325.
[71] S. A. Shah and F. Fioranelli, “Rf sensing technologies for assisted daily [Online]. Available: https://doi.org/10.1145/3307334.3326081
living in healthcare: A comprehensive review,” IEEE Aerospace and [90] J. Yang, H. Zou, Y. Zhou, and L. Xie, “Learning Gestures From WiFi:
Electronic Systems Magazine, vol. 34, no. 11, pp. 26–44, 2019. A Siamese Recurrent Convolutional Architecture,” IEEE Internet of
[72] H. Abdelnasser, M. Youssef, and K. A. Harras, “Wigest: A ubiquitous Things Journal, vol. 6, no. 6, pp. 10 763–10 772, 2019.
wifi-based gesture recognition system,” in 2015 IEEE Conference on [91] B. Huang, G. Mao, Y. Qin, and Y. Wei, “Pedestrian flow estimation
Computer Communications (INFOCOM), 2015, pp. 1472–1480. through passive wifi sensing,” IEEE Transactions on Mobile Computing,
[73] Q. Pu, S. Gupta, S. Gollakota, and S. Patel, “Gesture Recognition Using vol. 20, no. 4, pp. 1529–1542, 2021.
Wireless Signals,” GetMobile: Mobile Computing and Communications, [92] Y. Bai and X. Wang, “CARIN: Wireless CSI-based driver activity
vol. 18, no. 4, pp. 15–18, 2015. recognition under the interference of passengers,” Proceedings of the
[74] C. Wu, Z. Yang, Z. Zhou, X. Liu, Y. Liu, and J. Cao, “Non-invasive ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies,
detection of moving and stationary human with wifi,” IEEE Journal vol. 4, no. 1, 2020.
on Selected Areas in Communications, vol. 33, no. 11, pp. 2329–2342, [93] J. Yang, H. Zou, S. Cao, Z. Chen, and L. Xie, “MobileDA : Toward
2015. Edge-Domain Adaptation,” vol. 7, no. 8, pp. 6909–6918, 2020.
[75] M. Kotaru, K. Joshi, D. Bharadia, and S. Katti, “Spotfi: Decimeter [94] H. Fei, F. Xiao, J. Han, H. Huang, and L. Sun, “Multi-variations activity
level localization using wifi,” SIGCOMM Comput. Commun. Rev., based gaits recognition using commodity wifi,” IEEE Transactions on
vol. 45, no. 4, p. 269–282, Aug. 2015. [Online]. Available: Vehicular Technology, vol. 69, no. 2, pp. 2263–2273, 2020.
https://doi.org/10.1145/2829988.2787487 [95] Z. Zhou, Z. Yang, C. Wu, L. Shangguan, and Y. Liu, “Omnidirectional
[76] K. Joshi, D. Bharadia, M. Kotaru, and S. Katti, “Wideo: Fine-grained coverage for device-free passive human detection,” IEEE Transactions
device-free motion tracing using {RF} backscatter,” in 12th {USENIX} on Parallel and Distributed Systems, vol. 25, no. 7, pp. 1819–1829,
Symposium on Networked Systems Design and Implementation ({NSDI} 2014.
15), 2015, pp. 189–204. [96] K. Qian, C. Wu, Z. Yang, Y. Liu, and Z. Zhou, “PADS: Passive detection
[77] X. Li, S. Li, D. Zhang, J. Xiong, Y. Wang, and H. Mei, “Dynamic- of moving targets with dynamic speed using PHY layer information,”
music: Accurate device-free indoor localization,” in Proceedings of the Proceedings of the International Conference on Parallel and Distributed
2016 ACM International Joint Conference on Pervasive and Ubiquitous Systems - ICPADS, vol. 2015-April, pp. 1–8, 2014.
Computing, ser. UbiComp ’16. New York, NY, USA: Association [97] K. Ohara, T. Maekawa, and Y. Matsushita, “Detecting state changes of
for Computing Machinery, 2016, p. 196–207. [Online]. Available: indoor everyday objects using wi-fi channel state information,” Proc.
https://doi.org/10.1145/2971648.2971665 ACM Interact. Mob. Wearable Ubiquitous Technol., vol. 1, no. 3, Sep.
[78] Y. Xie, J. Xiong, M. Li, and K. Jamieson, “xD-Track: Leveraging multi- 2017. [Online]. Available: https://doi.org/10.1145/3131898
dimensional information for passive Wi-Fi tracking,” Proceedings of the [98] X. Wang, C. Yang, and S. Mao, “TensorBeat: Tensor decomposition for
Annual International Conference on Mobile Computing and Networking, monitoring multi-person breathing beats with commodity WiFi,” arXiv,
MOBICOM, pp. 39–43, 2016. vol. V, pp. 1–28, 2017.
[79] K. Qian, C. Wu, Z. Yang, C. Yang, and Y. Liu, “Decimeter level [99] M. M. Breunig, H. P. Kriegel, R. T. Ng, and J. Sander, “Lof: Identifying
passive tracking with WiFi,” in Proceedings of the Annual International density-based local outliers,” Proc.acm Sigmod Int.conf.on Management
Conference on Mobile Computing and Networking, MOBICOM, 2016, of Data, 2000.
pp. 44–48. [100] H. Wang, D. Zhang, J. Ma, Y. Wang, Y. Wang, D. Wu, T. Gu, and
[80] X. Zheng, J. Wang, L. Shangguan, Z. Zhou, and Y. Liu, “Smokey: B. Xie, “Human respiration detection with commodity wifi devices:
Ubiquitous smoking detection with commercial WiFi infrastructures,” do user location and body orientation matter?” in Proceedings of the
in Proceedings - IEEE INFOCOM, vol. 2016-July, no. Cv. IEEE, 2016 ACM International Joint Conference on Pervasive and Ubiquitous
2016. Computing, 2016, pp. 25–36.
[81] W. Wang, A. X. Liu, and M. Shahzad, “Gait recognition using WiFi [101] C. Ieracitano, A. Adeel, M. Gogate, K. Dashtipour, F. C. Morabito,
signals,” UbiComp 2016 - Proceedings of the 2016 ACM International H. Larijani, A. Raza, and A. Hussain, “Statistical analysis driven
Joint Conference on Pervasive and Ubiquitous Computing, pp. 363–373, optimized deep learning system for intrusion detection,” in International
2016. Conference on Brain Inspired Cognitive Systems. Springer, 2018, pp.
[82] H. Li, W. Yang, J. Wang, Y. Xu, and L. Huang, “WiFinger: Talk 759–769.
to your smart devices with finger-grained gesture,” UbiComp 2016 [102] P. Melgarejo, X. Zhang, P. Ramanathan, and D. Chu, “Leveraging
- Proceedings of the 2016 ACM International Joint Conference on directional antenna capabilities for fine-grained gesture recognition,”
Pervasive and Ubiquitous Computing, pp. 250–261, 2016. UbiComp 2014 - Proceedings of the 2014 ACM International Joint
[83] W. Wang, A. X. Liu, M. Shahzad, K. Ling, and S. Lu, “Device-Free Conference on Pervasive and Ubiquitous Computing, pp. 541–551,
Human Activity Recognition Using Commercial WiFi Devices,” IEEE 2014.
Journal on Selected Areas in Communications, vol. 35, no. 5, pp. 1118– [103] F. Wang, S. Panev, Z. Dai, J. Han, and D. Huang, “Can WiFi estimate
1131, 2017. person pose?” arXiv:1904.00277, no. March, 2019.
[84] K. Qian, C. Wu, Z. Yang, Y. Liu, and K. Jamieson, “Widar: Decimeter- [104] L. Guo, Z. Lu, X. Wen, S. Zhou, and Z. Han, “From signal to image:
level passive tracking via velocity monitoring with commodity Wi- Capturing fine-grained human poses with commodity wi-fi,” IEEE
Fi,” Proceedings of the International Symposium on Mobile Ad Hoc Communications Letters, vol. 24, no. 4, pp. 802–806, 2020.
Networking and Computing (MobiHoc), vol. Part F1291, 2017. [105] F. Li, M. Valero, H. Shahriar, R. A. Khan, and S. I. Ahamed,
[85] K. Qian, C. Wu, Z. Yang, Y. Liu, H. E. Fugui, and T. Xing, “Enabling “Wi-COVID: A COVID-19 symptom detection and patient monitoring
contactless detection of moving humans with dynamic speeds using framework using WiFi,” Smart Health, vol. 19, no. November 2020, p.
CSI,” ACM Transactions on Embedded Computing Systems, vol. 17, 100147, 2021. [Online]. Available: https://doi.org/10.1016/j.smhl.2020.
no. 2, pp. 1–18, 2018. 100147

19
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

[106] C. Uysal and T. Filik, “RF-Based noncontact respiratory rate monitoring

with parametric spectral estimation,” IEEE Sensors Journal, vol. 19, Ahmad Taha (MIEEE) is a Lecturer
no. 21, pp. 9841–9849, 2019.
[107] K. Qian, C. Wu, Y. Zhang, G. Zhang, Z. Yang, and Y. Liu, “Widar2.0: in the Glasgow College UESTC at the
Passive human tracking with a single Wi-Fi link,” MobiSys 2018 - University of Glasgow. He is Endorsed
Proceedings of the 16th ACM International Conference on Mobile by the Royal Academy of Engineering
Systems, Applications, and Services, pp. 350–361, 2018.
[108] C. Lowanichkiattikul, M. Dhanachai, C. Sitathanee, S. Khachonkham, as an Exceptional Promise under the
and P. Khaothong, “Impact of chest wall motion caused by respiration Global Talent scheme, a Fellow of Ad-
in adjuvant radiotherapy for postoperative breast cancer patients,” vanced Higher Education (FHEA), and
SpringerPlus, vol. 5, no. 1, pp. 1–8, 2016.
[109] Y. Zhang, Y. Zheng, G. Zhang, K. Qian, C. Qian, and Z. Yang, “Gaitid:
a UKCGE recognised Associate Super-
Robust wi-fi based gait recognition,” in International Conference on visor. He first joined the University of
Wireless Algorithms, Systems, and Applications. Springer, 2020, pp. Glasgow in October 2020 as a Postdoc-
730–742.
[110] T. Starner, J. Auxier, D. Ashbrook, and M. Gandy, “The gesture pendant:
toral Research Associate working in collaboration with several
a self-illuminating, wearable, infrared computer vision system for home industrial and academic partners including Cisco and the
automation control and medical monitoring,” in Digest of Papers. Fourth University of Strathclyde, on the 5G New Thinking project,
International Symposium on Wearable Computers, 2000, pp. 87–94.
[111] F. Zhou, T. Yu, R. Du, G. Fan, Y. Liu, Z. Liu, J. Xiang, Y. Wang,
funded by the Department for Digital, Culture, Media and
B. Song, X. Gu, L. Guan, Y. Wei, H. Li, X. Wu, J. Xu, S. Tu, Y. Zhang, Sport (DCMS). He has 8+ years’ experience working in Higher
H. Chen, and B. Cao, “Clinical course and risk factors for mortality Education (HE) Institutions, delivering teaching in Egypt, the
of adult inpatients with COVID-19 in Wuhan, China: a retrospective
cohort study,” The Lancet, vol. 395, no. 10229, pp. 1054–1062, 2020.
UK and China.
[Online]. Available: http://dx.doi.org/10.1016/S0140-6736(20)30566-3 He received a dual B.Sc. with Honours from October
[112] W.-j. Guan, Z.-y. Ni, Y. Hu, W.-h. Liang, C.-q. Ou, J.-x. He, L. Liu, University for Modern Sciences and Arts (MSA) in Egypt and
H. Shan, C.-l. Lei, D. S. Hui, B. Du, L.-j. Li, G. Zeng, K.-Y. the University of Greenwich in the UK, in 2012. He was a
Yuen, R.-c. Chen, C.-l. Tang, T. Wang, P.-y. Chen, J. Xiang, S.-y.
Li, J.-l. Wang, Z.-j. Liang, Y.-x. Peng, L. Wei, Y. Liu, Y.-h. Hu, recipient of two scholarships to pursue his M.Sc. and PhD,
P. Peng, J.-m. Wang, J.-y. Liu, Z. Chen, G. Li, Z.-j. Zheng, S.-q. Qiu, after a successful Vice-Chancellor award application, in 2013
J. Luo, C.-j. Ye, S.-y. Zhu, and N.-s. Zhong, “Clinical characteristics and 2016, respectively, at the University of Greenwich. He
of coronavirus disease 2019 in china,” New England Journal of
Medicine, vol. 382, no. 18, pp. 1708–1720, 2020. [Online]. Available: completed his M.Sc degree, with a distinction, in Embedded
https://doi.org/10.1056/NEJMoa2002032 Systems in 2014 and his PhD in 2020, which was partially
[113] Z. Yang, C. Wu, and Y. Liu, “Locating in fingerprint space: Wireless funded and in collaboration with Medway NHS Foundation
indoor localization with little human intervention,” Proceedings of the
Annual International Conference on Mobile Computing and Networking, Trust in Kent, UK. He was nominated for the Energy Institute
MOBICOM, pp. 269–280, 2012. award in 2019 due to contributions in technology-based
[114] S. Shi, Y. Xie, M. Li, A. X. Liu, and J. Zhao, “Synthesizing wider energy-saving systems in the NHS.
wifi bandwidth for respiration rate monitoring in dynamic environ-
ments,” in IEEE INFOCOM 2019 - IEEE Conference on Computer
Communications, 2019, pp. 181–189. Syed Aziz Shah is an Associate Pro-
[115] Y. Zhang, Y. Zheng, G. Zhang, K. Qian, C. Qian, and Z. Yang, “Gaitid: fessor of Mobile Health at the Healthcare
Robust wi-fi based gait recognition,” in Wireless Algorithms, Systems,
and Applications, D. Yu, F. Dressler, and J. Yu, Eds. Cham: Springer Technology and Innovation Theme, Re-
International Publishing, 2020, pp. 730–742. search Centre for Intelligent Healthcare.
[116] B. Zheng and R. Zhang, “Intelligent reflecting surface-enhanced Syed is working on software defined
ofdm: Channel estimation and reflection optimization,” IEEE Wireless
Communications Letters, vol. 9, no. 4, pp. 518–522, 2020. radios to develop a prototype that can
[117] S. Liaqat, K. Dashtipour, S. A. Shah, A. Rizwan, A. A. Alotaibi, monitor activities of daily livings of
T. Althobaiti, K. Arshad, K. Assaleh, and N. Ramzan, “Novel ensemble elderly people and detect anomalies in
algorithm for multiple activity recognition in elderly people exploiting
ubiquitous sensing devices,” IEEE Sensors Journal, vol. 21, no. 16, pp.
respiratory rate at homes or care-centres.
18 214–18 221, 2021. Dr. Aziz’s research work spans across multiple disciplines
[118] S. Imran, T. Mahmood, A. Morshed, and T. Sellis, “Big data analytics including wireless sensing, radar technology, software de-
in healthcare a systematic literature review and roadmap for practical
implementation,” IEEE/CAA Journal of Automatica Sinica, vol. 8, no. 1,
fined radios, machine learning, cyber security and intelligent
pp. 1–22, 2021. healthcare technologies. Syed has (co) authored more than 60
[119] L. Zhang, H. Han, M. C. Zhou, Y. Al-Turki, and A. Abusorrah, “An technical articles in top-rank peer-reviewed multi-disciplinary
improved discriminative model prediction approach to real-time tracking journals (3 transactions) focusing on intelligent healthcare. In
of objects with camera as sensors,” IEEE Sensors Journal, 2021.
[120] K. Qian, Z. He, and X. Zhang, “3D Point Cloud Generation with addition, Syed has taught numerous graduate and postgraduate
Millimeter-Wave Radar,” Proceedings of the ACM on Interactive, Mobile, modules and supervised students at higher education level.
Wearable and Ubiquitous Technologies, vol. 4, no. 4, 2020. His research interests include but not limited to mobile
Yao Ge received the dual B.Eng. health, prototype design, radar sensing for healthcare
degree in Electrical and Electricity technologies, non-invasive fall detection, physiological
Engineering from UESTC, China and measurements, remote patient monitoring wireless sensing,
University of Glasgow with first honours, machine learning and cyber security for intelligent healthcare.
in 2020. He is currently pursuing his
Ph.D. degree in University of Glasgow,
U.K. His research interests include
wireless sensing, artificial intelligence,
and smart healthcare systems.

20
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/RBME.2022.3156810, IEEE Reviews
in Biomedical Engineering

Kia Dashtipour received the M.Sc. research including URSI 2019 Young Scientist Awards,
degree in Advanced Computer System UK exceptional talent endorsement by Royal Academy of
Development in 2014 and Ph.D. degree Engineering, Sensor 2021 Young Scientist Award , National
in computing science in 2019 from talent pool award by Pakistan, International Young Scientist
University of Stirling. He is currently a Award by NSFC China, National interest waiver by USA
lecturer at Edinburgh Napier University. and 8 best paper awards. He is a committee member for
He was previously research associate at IEEE APS Young professional, Sub-committee chair for
the University of Glasgow from 2019 to IEEE YP Ambassador program, IEEE 1906.1.1 standard on
2021. His main research interests include natural language nano communication, IEEE APS/SC WG P145, IET Antenna
processing, IoT, machine learning and speech enhancement. Propagation and healthcare network.

Shuyuan Zhu (S’08-A’09-M’13) Muhammad Ali Imran (M’03,

received the Ph.D. degree from the SM’12) Fellow IET, Senior Member
Hong Kong University of Science and IEEE, Senior Fellow HEA is Dean
Technology (HKUST), Hong Kong, in University of Glasgow UESTC and a
2010. From 2010 to 2012, he worked Professor of Wireless Communication
at HKUST and Hong Kong Applied Systems with research interests in self
Science and Technology Research organised networks, wireless networked
Institute Company Limited, respectively. control systems, internet of things (IoT)
In 2013, he joined University of and the wireless sensor systems. He
Electronic Science and Technology of heads the Communications, Sensing and Imaging (CSI)
China and is currently a Professor with research group at University of Glasgow and is the Director
School of Information and Communication Engineering. Dr. of Glasgow-UESTC Centre for Educational Development and
Zhu’s research interests include image/video compression Innovation. He is an Affiliate Professor at the University of
and image processing. He received the Top 10% paper Oklahoma, USA and a visiting Professor at 5G Innovation
award at IEEE ICIP-2014 and the Best 10% paper award Centre, University of Surrey, UK. He has over 20 years of
at VCIP-2016. He currently serves as an Associate Editor combined academic and industry experience with several
of IEEE Transactions on Circuits and Systems for Video leading roles in multi-million pounds funded projects. He has
Technology. Dr. Zhu was the session chairs of IEEE DSP-2015, filed 15 patents; has authored/co-authored over 400 journal
IEEE VCIP-2020, IEEE ICIP-2021 and is also the session and conference publications; has edited 7 books and authored
chair of IEEE ICME-2022. He served as the committee more than 30 book chapters; has successfully supervised over
members for IEEE ICME-2014, VCIP-2016, PCM-2017 and 40 postgraduate students at Doctoral level. He has been a
IEEE MIPR-2020. He is a member of IEEE CAS society. consultant to international projects and local companies in the
area of self-organised networks.
Jon Cooper holds The Wolfson
Chair in Biomedical Engineering and is
Emeritus Vice Principal at the University
of Glasgow. His major research interests
are in medical diagnostics and imaging,
with a track record of spin-out and
translation of devices into industry and
practice. He was elected as a Fellow
of the Royal Academy of Engineering
(UK’s national academy of engineering) as well as a Fellow of
the Royal Society of Edinburgh (Scotland’s national academy
of arts, humanities and sciences). He has published over 260
paper, 18 books and book chapters and has an H-index of 48.

Qammer H. Abbasi (SMIEEE,

MIET, FRET, FRSA), Dr Abbasi is a
Reader with the James Watt School
of Engineering, University of Glasgow,
U.K., deputy head for Communication
Sensing and Imaging group. He has
published 350+ leading international
technical journal and peer reviewed
conference papers and 10 books and
received several recognitions for his

21
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/