International Journal of Engineering & Technology, 7 (3.

24) (2018) 505-511

International Journal of Engineering & Technology

Website: www.sciencepubco.com/index.php/IJET

Research paper

Online Analysis of Handwriting for Disease Diagnosis: A Review

1*
Seema Kedar, 2D. S. Bormane, 3Sandeep Joshi
1
Researcher, RSCOE, Savitribai Pule Pune University, Pune, India, 411033.
2
Professor, AISSMSCOE, Savitribai Pule Pune University, Pune, India,411001.
3
Researcher, RSCOE, Savitribai Pule Pune University, Pune, India, 411033.
*
Corresponding author E-mail:seema_kedar@yahoo.com

Abstract

Background/Objectives: Handwriting is an action governed by brain like any other action. This process is usually unconscious and is
closely tied to impulses from brain. Any kind of disease affects the kinetic movement and reflects in subject’s handwriting. To understand
the health and mental problems, it is important to focus on how subject writes instead of what subject writes. This also makes the process
of handwriting analysis independent of any language. Handwriting analysis is a pseudo-science used to study physical and behavioral
characteristics of handwriting. In this paper, the general approach used for the disease diagnosis based on digital handwriting analysis has
been presented. The research work carried out to diagnose diseases such as Alzheimer, Mild Cognitive Impairment, Dysgraphia,
Schizophrenia, Autism, Parkinson’s disease and Mental illness based on digital handwriting analysis has been reviewed in this paper. The
features related to motion, time and pressure have been used for diagnosis of disease. The experiments and results are also summarized in
this paper.

Keywords: Handwriting Analysis, Handwriting Features, Tablet, Disease.

13. Pressure applied on writing organ while writing,

1. Introduction 14. Accent Marks and Periods which reflect memory disorders,
imagination and attention [4, 5].
All actions including writing start in the brain. Like all other The Graphologists make use of combination of two or more
movements, the act of writing depends on central nerve system features mentioned above for handwriting analysis. These features
[1]. Our brain sends impulses to hand through nervous signals, are extracted and converted into numeric values for statistical
achieving the motor act. Graphology is pseudo-science based on analysis and disease diagnosis.
combination of psychoanalysis and neuroscience nested in As shown in Figure 1, the online system used for handwriting
subconscious mind. Though handwriting is driven through pen, analysis of patients comprises below three components:
it’s movement is governed by the central nervous system, which is  Patient: Patient is a suspect having disease or mental illness.
a process usually unconscious, but most revealing [3].  Digitizing tablet: It is a device used to capture signals from
Handwriting is closely tied to impulses from the brain and patient’s handwriting.
therefore it can be reliably used to predict state of physical,  Computer: The computer with general configuration used to
emotional and mental health of individual [2]. Handwriting run image acquisition software and classification engine to
analysis is used to find out disturbance in the subject’s capture measurements of handwriting and to identify kind of
handwriting. disease or mental illness.
The important handwriting features used for disease diagnosis are:
1. Congestion: It is shown by letters having ovals and curls full
of ink,
2. Fragmentation: It is shown by disconnected curves of letters,
3. Direction of lines
4. Layout of Anomalies
5. Torsion: It is an irregularity or luxuriating of part of a letter
or entire letter,
6. Viscosity: It is unclear or dirty extension of upper and lower
parts of letters,
7. Shakiness: It is small disruptions in strokes of letters, Figure 1:. Components of online system used for handwriting analysis of
8. Slant: It is an uneven inclined right movement of pen on patients
paper while drawing letters,
9. Movement between stokes, 2. Approaches for Handwriting Analysis
10. Variation in size of letters while writing letters,
11. Alterations in shape of curves for similar letters, Two approaches used for handwriting analysis are explained
12. Breeze: It is the part of stroke over sheet paper, when pen below.
went without leaving ink,

Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original work is properly cited.
506 International Journal of Engineering & Technology

2.1. Offline Handwriting Analysis

The traditional methods of handwriting analysis are based on text

written on paper which is called as off-line handwriting. These
methods include Clock Drawing Test (CDT) [6], Mini Mental
State Examination (MMSE test) [7], House-Tree-Person (HTP)
test [8]. In CDT, subjects have been asked to draw a clock with all
digits pointing hands to 11:50. Alzheimer Patients face issues in
placing digits; the clock digits may not evenly have been spaced
or may show incorrect time. Scores are measured on scale of 15
points, for shape and spatial arrangement of clock digits and
hands. MMSE measures cognitive functions related to attention,
language, registration, recall, calculation, orientation, and ability
to follow simple commands. The test has 11 questions with
maximum score of 30-points. In HTP test patients have been
asked to draw a tree, a house and a person. Along with drawing
tasks, few questions are introduced to understand personality. The
test may be extended to evaluate brain damage.

2.2. Online Handwriting Analysis

With evolution of digital world, graphologists started using digital

tablet which provided richer set of measures on handwriting called Figure 2: Generalized machine learning approach for online handwriting
as online handwriting. The online analysis of handwriting finds analysis [19].
temporal features such as inclination, time-stamp and pressure
applied on pen, and velocity of the pen movements, which is not As shown in Figure 3, the tasks include:
possible to capture in off-line handwriting [9]. However, on-line 1) Draw and copy two pentagons based on MMSE test
handwriting introduces distortion and transition trajectories 2) Draw a House Tree Person based on HTP test
(curves drawn for continuous or overlapped writing), which need 3) Write four provided words in capital letters
to be removed before processing. Various classification and
4) Draw a clock pointing hands to specific time based
clustering techniques have been used to find out the best features
suited for the analysis of targeted disease. on CDT test
5) Copy a phonetically complete sentence in cursive
letters
3. Generalized Approach for Online
Handwriting Analysis
The generalized machine learning approach for online handwriting
analysis is shown in Figure 2. It consists of two phases namely
learning phase and testing phase. The machine learning model is
used as it improves classification accuracy based on experience.
The process of digital handwriting analysis begins with collecting
samples of patients of targeted disease. Subjects having similar
age group and same kind of health or mental illness are chosen to
train classifier for targeted disease. This phase is called learning
phase. The classifier is a model built using knowledge gained
from various measures of handwriting. This model is used in
testing phase for disease diagnosis.

The various steps of generalized machine learning approach are:

a. Data Acquisition: Figure 3: Writing-Drawing Tasks [3]
The data have been recorded using INTUOS WACOM series 4
digitizing tablet with INTUOS Ink-pen. The device has ability to To reduce distortion while using digital device for handwriting, a
capture spatial movements and pressure applied on tablet. It also normal A4 size paper is laid on the tablet and subject has been
captures pen tip movement even when it is not rested on tablet’s asked to write on paper. Using signals recorded by tablet, an
surface up-to certain height (6 mm). This feature allows to image of handwriting can be viewed on the screen of computer
measure in-air handwriting movements. Five tasks related to connected to it. Based on these signals acquisition software
writing/drawing are given to subjects with proper instructions. generates a .svc file which has been used for statistical analysis
These tasks are based on, drawing and writing tasks well practiced [10]. The .svc file is an ASCII file which can be read using any
in clinical diagnosis of mental disorders. From these tasks the standard editor application like WORD, NOTEPAD. As shown in
measurements are extracted. Figure 4, the .svc file contains following information:
 Time stamp
 Position on x-axis
 Position on y-axis
 Pen status (up=0 or down =1)
 Pressure applied on pen
 Pen’s Azimuth angle with respect to tablet
 Pen’s Altitude angle with respect to tablet
507 International Journal of Engineering & Technology

There are two phases of system namely learning phase and testing
phase. The goal of learning phase is to build the model of
classification. Machine learning algorithms are used to build this
model [12]. The ANOVA test [13], Naive Bayes [14], Support
Vector Machine (SVM), Decision Tree, K-Nearest Neighbors, K-
Means [15] are most commonly used algorithms to model the
data.

E. Revalidation and Correction

The model of classification formed above needs to be revalidated
and corrective actions need to be taken to improve accuracy of
results. In learning phase, the first selection of features, made for
building the model, is random and this may cause reduced
accuracy of results. To reduce the impact of randomness, one
Figure 4: Contents of .svc file [3]
feature is chosen at a time which has not been selected yet (called
Out-of-Bag, OOB, data) [16] and the accuracy obtained by it is
b. Segmentation
compared with previous model’s accuracy. The feature producing
To identify behavior, the handwriting trajectory is divided into
short-length continuous segments. Segmentation improves the maximum accuracy is selected first and model is rebuilt. The
results can be validated by comparing results obtained with
efficiency, as it helps to identify common patterns of handwriting.
clinical tests which are not part of learning phase.
Segmentation also removes outliers introduced while capturing
handwriting using digital tablet. The set of measurements for a
segment forms a feature which is used to train classifier in 4. Disease and Related Online Handwriting
learning and to derive results in testing phase. The image obtained Analysis
from the INTUOS WACOM digitalizing tablet has series of
alternate in-air and on-paper curves. The segmentation has been
The research work carried out to diagnose various critical diseases
done easily by monitoring .svc file’s Pen Status field. An arc
through online handwriting analysis is explained below.
between adjacent changes in Pen’s status forms segment.

C. Feature Extraction 4.1. Alzheimer and Mild Cognitive Impairment Disease

Researchers have analyzed handwriting with wide range of
features. Below is the list of features that are used for analysis of Alzheimer disease is found in about 10% of people with around 65
online handwriting: years’ age. Loss of brain cells causes this disease and directly
1) Number of on-paper and in-air strokes while completing impacts the cognitive functions of human. Mild Cognitive
a task. Impairment (MCI) is an intermediate stage of Alzheimer Disease
2) Time to complete entire task. (AD) in which people starts losing memory and thinking abilities.
3) Total time spent in-air for completing a task. Patients with high degree of Alzheimer are hardly able to put pen
4) Total time spent on-paper for completing a task. on paper. Alzheimer and MCI patients show significant difference
5) Width, height, orientation (slope) and length of segment. in kinetic and pressure features of handwriting as compared to
6) Mean distance between letters of a segment. healthy people [17].
7) Mean distance between consecutive letters of same word Gady Bar-Onet et. al. examined the hand-writing process of
of a segment. persons having MCI and Alzheimer’s disease. They used spatial,
8) Mean pressure applied on pen. temporal and pressure measures as primary features of
9) Standard deviation in pressure applied on pen. handwriting. The spatial features included total length of on-paper
10) Speed of trajectory calculated as length of segment strokes measured in millimeter. Temporal features included time
divided by duration. taken to finish the task, time spent in-air and on-paper measured in
11) Velocity: It is the rate at which the position of a pen seconds. They also worked on other important feature namely the
changes with time. mean velocity of pen when it was on paper. The mean pressure
12) Acceleration: It is the rate at which the velocity of a pen applied by pen measured with non-scaled units from 0 to 1024,
changes with time. was also included. The results showed continuous variation in in-
13) Jerk: It is the rate at which the acceleration of a pen air time of normal people and patients. The in-air duration was
changes with time. significantly greater for subjects with MCI and Mild AD as
14) Median of power spectral density which takes into compared to healthy normal participants. Except velocity, all
account a distribution of energy in power spectrum. kinematic and pressure measures consistently varied between Mild
15) Lamed height ratio of the segment (Letters raised AD subjects and normal healthy people. Healthy normal people
considerably above the line called Lamed letter). applied significantly more pressure on pen in comparison with
16) Deletions i.e. the number of letters or letter segments Mild AD and MCI subjects. The kinematic features showed 10%
crossing each other. to 27% variation in strength of relationship (ETA- Squared scores)
among scores of group of MCI, Mild AD and normal subjects
While extracting feature, it is important to consider age difference [18].
between patients, distortion induced by using digital media while Fontana Paola et. al. devised a score system to analyze
recording handwriting and importance of one feature over the handwriting, and to test its relation with mental deterioration. The
other [11]. After measuring one or more features mentioned above score system had good inter-rater reliability and significantly
for each segment, various sampling and statistical techniques are correlates with Milan Overall Dementia Assessment scale
applied to improve quality of data. Sampling techniques include (MODA) and Mini Mental State Examination (MMSE). They
choosing segments of specific length or time. To obtain feature studied 25 subjects suspected to have mental disorder referred by
value, the statistical techniques like mean, median, average, Neuropsychology Service. All patients went through MODA and
standard deviation (SD) or root mean square are applied on MMSE test to measure cognitive disorder. To measure
sampled data. handwriting features, subjects were asked to write a text of six to
D. Building Model for Classification seven lines of their choice dictated by the examiner on a boxed
508 International Journal of Engineering & Technology

sheet and on a blank sheet. The statistical study was carried out features were classified using linear support vector machine. They
based on orientation, slant, ordering, and baseline of letters to obtained 89.9% accuracy [24].
derive writing score. This score was obtained using interrater
reliability coefficient of statistical program SPSS and the median 4.3. Schizophrenia
score of the three separate evaluations computed for each subject.
The results showed that patients having a writing score upto five Schizophrenia (SZ) is found in 1% of population in people of age
never had a MODA greater than 60 nor a MMSE greater than 20 around 20’s. Schizophrenia patients often behave quite different
[19]. They concluded that handwriting is a useful for evaluating and strange as compared to normal people and sometimes they
the posthumous of testamentary capacity. could not able to relate themselves with real world. Schizophrenia
Murad Badarna et. al. studied pen motion patterns in patients patients can be distinguished by using kinetic features of
having low MCI and high MCI. The handwriting analysis was handwriting like velocity or jerk [25].
based on spatial measures like width, height, displacement, and Alexander B et. al. studied Schizophrenia patient’s handwriting
curve length, kinematic measures like mean, median, and standard movements to measure drug-induced motor side effects. Six
deviation in acceleration, in-air and on-paper duration, pressure schizophrenia subjects who have not received any antipsychotic
and angles. They obtained classification accuracy of 95.23% for medication, 27 schizophrenia subjects treated with risperidone and
patients with low MCI based on the on-surface pen motion pattern 46 healthy subjects were enrolled for study. Participants were
and accuracy of 85.80% for patients having high MCI [20]. given exercise to draw loops of different size and writing a
sentence. The data was recorded and analysed using MovAlyzeR
4.2. Dysgraphia tool. The up and down successive movements were segmented
into strokes. For each stroke, peak vertical velocity, velocity
Dysgraphia is found in school-aged children. Children with scaling (VS), vertical size, and average normalized jerk (ANJ)
Dysgraphia face difficulties in organizing letters and produce across a trial were measured. Results showed that patients treated
uneven patterns of disordered handwriting. Using spatial and with risperidone exhibited significantly greater dysfluent
temporal measures, children with Dysgraphia can be identified handwriting movements than healthy or untreated SZ participants.
easily [21]. Differences were observed in some handwriting kinematic
Zdenek Mzourek et. al. analyzed sequential writing of 27 children measures like ANJ between unmediated SZ patients and healthy
having Dysgraphia, with 27 age-matched controls having participants. The sentences produced by unmediated SZ patients
experience of two years of writing in school. The parameters were were smoother than medicated patients, and their handwriting
based on 51 features, divided into three groups: non-linear movements were more dysfluent [26].
dynamic features, kinematic features and other features. The M. Ahmadlou et. al. performed analysis of kinetic features of
observations were based on altitude/tilt and pressure. The features handwriting among group of healthy people and SZ patients.
based on pressure found more useful for diagnosis of Dysgraphia. Subjects were asked to write ‘hello’ word three times in different
They observed that pressure applied was higher by 17% in case of scales of height 1cm, 2cm and 4cm. Eight different features were
children having Dysgraphia. To improve accuracy, they proposed extracted out of which three related to vertical peak velocity (PV)
an intra-writer normalization method based on subtraction. This and others are related to velocity scaling slope (VSS), normalized
method showed increase in accuracy by 4% and decrease in HPSQ jerk (NJ) and ANJ. An artificial neural network (ANN) of N nodes
score estimation error by 3.48%. This proposed automatic system (N equals to number of selected features) was created with three
had 96% sensitivity and specificity towards diagnosis of layers. The first layer was an input layer, next was hidden layer
Dysgraphia and also the system was able to rate developmental with 1 to 20 nodes and output layer with one node. The results
Dysgraphia with estimated HPSQ total score with 10% error [22]. were obtained through 100 different runs. In each run, four data
Patrice L. Weiss et. al. studied and compared both, online samples were used for testing purpose and other seventeen for
handwriting obtained using digital tablet and traditional learning purpose. The system was able to differentiate SZ subjects
handwriting analysis obtained from a pen and paper, among group from normal people with accuracy of 97.5% [27].
of children having Dysgraphia and group of children with Theo Wayne S. Fenton et. al. suggested a method to analyze
proficient handwriting. Handwriting samples of 50 proficient performance of lateralized motor in SZ subjects. Patients were
students and 50 students of 3rd grade with Dysgraphia were asked to draw two straight lines with each hand. These lines were
collected. The results showed that both digital and conventional scanned using HP Plotter. To measure deviation in straight line,
evaluations were able to differentiate between children with root mean squared (RMS) error of regression equation was
Dysgraphia and children with proficient handwriting. By calculated for each line. To measure overall degree of disorder, the
combining both the methods, they got improved understanding of average of RMS error of all four lines was taken. Also the
writing difficulties. The spatial and temporal measures of difference in RMS values was measured between two hands to
handwriting kinematics were used as primary features. The total compare performance index of motoric laterality. The writing
length paragraph and total length of in-air and on-paper movement samples were collected from 86 SZ patients diagnosed using
constitute spatial measures whereas total time recorded to finish statistical and diagnostic manual of mental disorders test. Fourteen
writing tasks, in-air and on-paper time, were also included as employees from housekeeping staff were included as control
temporal measures. The digital analysis of handwriting showed subjects. To obtain reliable image of scanned drawing sample,
significantly higher in-air time among the Dysgraphia handwrites four lines drawn by each control subject were selected randomly
as compared to proficient writers. The pencil travelled above from database and scanned twice and scores were compared using
writing surface between successive character segments, letters and one-tailed Pearson Correlation for each line drawn by each control
words among the non-proficient writers as compared to proficient subject. The SZ patients and control subjects were compared
writers. These results provided clues of underlying difficulties based on lateralization and motor disorder indices with the help of
limiting performance of children with Dysgraphia [23]. one-tailed T-tests. The SZ patients poorly performed motor
Gideon Dror et. al. proposed a system for characterization and movements than control subjects [28].
automatic identification of Dysgraphia in third-grade children. The
system was based on analysis of child’s writing movements. The 4.4. Autism
pressure applied on pen, it’s orientation and position were
measured using a standard digital writing pad. The samples of Autism occurs in school-aged children. Children with Autism
ninety-nine Dysgraphia writers and proficient writers were often face difficulties in communicating with other people. Also it
collected. The features extracted were based on wide range of is difficult to understand their feelings and read their mind.
dynamic and visual properties of handwriting. The extracted Children with Autism can be identified by analyzing speed, slant
509 International Journal of Engineering & Technology

and size of handwriting [29]. lengths, 9.22mm (SD = 1.60) whereas in positive and neutral
Sara Rosenblum studied behavior of children’s handwriting with subjects they were longer 10.67mm (SD = 2.46) and 10.74mm
autism spectrum disorder. Sixty children aged 9-12 years from (SD = 2.89) respectively. The width of strokes was narrower in the
third to sixth grades from different schools were included in the negative subjects, 3.03mm (SD = 0.44) as compared to 3.60mm
study. Out of sixty, half children had high-functioning autism (SD = 0.85) and 3.43mm (SD = 0.77) in positive and neutral
spectrum disorder having Intelligence Quotient above 80 whereas subjects respectively. In group of negative subjects, the height of
other 30 were normal children. The degree of pressure applied by strokes produced was shorter 4.52mm (SD = 0.9) in comparison
subject on paper, the rhythm and speed of handwriting, in-air time, with negative and neutral subjects 5.32mm (SD = 1.34) and
slant and some other features were measured for each subject. The 5.53mm (SD = 1.41) respectively. The duration of on-paper
system correctly identified children with autism with accuracy of strokes was shorter in the positive and negative moods, 144ms
91.5%. The letters produced by children with Autism were taller (SD = 19.2) and 143.5ms (SD = 18.4) respectively, whereas for
and broader, and degree of pen’s slant on paper was smaller, also neutral subjects it was 157 ms (SD = 32.5). Also strokes produced
on-paper and in-air waiting times were longer as compared to by subjects with negative mood had lower pressure than those
normal students [30]. made in neutral or positive moods [36].
A study by Beth Patricia Johnson et. al. was aimed to characterize Perla Werner et. al. examined functional disorder in performance
handwriting performance in children with Autism Spectrum of handwriting between elder people with Mild Major Depressive
Disorder(ASD) by using digitized handwriting tasks. The study Disorder (MDD). Twenty elder people diagnosed with DSM-IV
included 52 boys with 8 to 12 years of age, 29 typically test of mild MDD were included along with twenty healthy
developing (TD) children from control group and 23 patients of subjects recruited among MDD participant’s relatives having
ASD. The efficiency of motor movements was measured using similar gender, age and educational level. The space, time and
MABC-2 test. Five writing sequences were given to each subject pressure were measured based on four handwriting tasks such as
in cursive letters (1. emem, 2. elel, 3. eeem, 4. eeel, 5. eeee) copying a paragraph, writing one’s name and surname, writing all
having transition movements among similar (e-to-e) and different alphabets sequentially and filling in a check. The pressure applied
letters (e-to-m, e-to-l) with variation in direction and sizing. by depressed patients was significantly lower and proven to be
Tortuosity, size, writing speed and average peak velocity for a major factor affecting the writing performance. Also the time
writing sequence were measured. The results showed that ASD required to plan and execute movement of hands was higher. This
children were facing greater difficulty in writing as compared to feature was measured using in-air duration on pen [37].
group of TD children of similar age. Also the letters/words were
spaced unevenly by ASD children [31]. 4.6. Parkinson Disease
A review by Evdokia Anagnostou et. al. analyzed existing
evidences showing dysfunction in writing movements of children Parkinson Disease (PD) is a disorder in motor functions due to
with ASD and documented their handwriting difficulties. They damage in central nervous system. It is found in people with age
collected and analysed data from January 1943 to January 2011 around 50 years and it gets worse over the time. By measuring the
from ISI Web of Science, PubMed, Google Scholar, Scopus and jerk and acceleration pattern of handwriting produced PD can be
all English-language related to ASD children [32]. identified [38].
The Quantitative measurement of handwriting was carried out by
4.5. Mental Illness David Song et. al. to analyse discriminate in every-day
handwriting tasks among group of people having Parkinson’s
Handwriting Analysis has also been used to find people with disease, people with psychotropic-induced Parkinsonism and
negative emotions like Stress, Anxiety and Depression. normal people. The aim was to find some psycho-parameter to
Depression is increasingly affecting people world-wide. It affects measure degree of Parkinsonism. The experiment included ten
motivation, concentration and many other aspects of human life. subjects (1 female and 9 males) meeting DSM-IV criteria of
Stress depends on two things: psychological perception of schizophrenia having clinically observable drug-induced
pressure and body’s response to it which involves multiple Parkinsonism, thirteen patients (4 females and 9 males) diagnosed
systems, from metabolism to muscles to memory. Anxiety is with idiopathic PD and Twelve (2 females and 10 males) normal
reaction to stressful situations which occurs in threatening healthy comparison subjects (NC). The velocity analysis of
circumstances. Person with negative emotions show difference in handwriting samples showed that peak velocities of vertical
pressure applied on pen while writing, also the strokes are strokes from middle to bottom were nearly doubled, whereas for
abnormal in time and space [33, 34]. SZ patient’s velocities were constant. In addition, to find degree of
Anna Esposito et. al. proposed a system to detect of negative smoothness of handwriting, NJ was calculated based on
emotions through writing and drawing tasks. The time-based normalized size and duration of strokes. Results showed that both
features were used as they are more efficient to determine the state SZ and PD patients had lower velocities than normal people,
of mental illness. They measured in-air and on-paper duration, particularly for stokes larger than 4cm. Significant reduction in VS
total time and various characteristics of stokes using digitalizing was observed in both PD and SZ subjects whereas on SZ subjects
tablet. For each writing or drawing task, in-air, on-paper, and total exhibited lack of smoothness [39].
time taken by candidate was recorded along with number of stokes Rifat Sipahi et. al. introduced method based on analysis of static
and ranked using Random Forest approach. The results were images of subject’s signature or handwriting samples to detect
validated against DASS [35] and repeating the experiment N times changes in micrographic of subjects with therapeutic response or
on K candidates by excluding one candidate from population each symptomatic progression in Parkinson’s disease. Twelve samples
time to form subset of K-1 candidates. The results showed that of signature of different subjects corresponding to symptomatic
stress and anxiety can be recognized more correctly than PD conditions, normal health and artificially generated signature
depression using digital handwriting analysis [11]. with reduced size were used for comparison. Matrices sensitive to
Gil Luria et. al. studied relationship between handwriting and the properties of micrographic were chosen with minimal
mood. Sixty-two subjects from University of Haifa were included sensitivity to confounding handwriting properties. These metrics
in the study with average age of 24.8 ranging between 21 to 29 included ink utilization, character size-reduction and pixel density
years. The Computerized Penmanship Evaluation Tool and of writing samples from left to right. The signatures of subjects
Software (ComPET) was used for analysis of features. The were collected and scored for each group before and after clinical
variation in length, width, height of strokes, duration and pressure diagnosis of subjects. Significant difference was observed in Pixel
applied were measured among group of positive, negative and Density among signature recordings, before versus after [40].
neutral people. In negative subjects the strokes had shorter M. Naumann et. al. studied kinetic features of handwriting
510 International Journal of Engineering & Technology

movements, speed, stroke duration, size and acceleration of [7] MMSE Test, Available from:
Parkinson’s disease subjects. For analysis, subjects were asked to https://www.healthdirect.gov.au/mini-mental-state-examination-
draw combination of letter ‘ll’ of the German words ‘helles’ mmse
[8] House Tree and Person test, Available from:
(bright) and ‘grelles’ (glaring). For evaluation, distance of writing
https://healthfully.com/interpret-housetreeperson-test-8631546.html
traces (in mm) of letter having ‘ll’ combination and maximum [9] Xu-Yao Zhang, Guo-Sen Xie, Cheng-Lin Liu, and Yoshua Bengio,
positive/negative absolute acceleration (slowing down) were “End-to-End Online Writer Identification with Recurrent Neural
measured in both descending and ascending strokes recorded for Network”, IEEE Transactions on Human-Machine Systems, vol.
each trial. Total writing time (in ms) and maximum/minimum 47, Issue: 2, 2017.
absolute (tangential) velocities were also calculated for [10] Andrew Seniuk and Dorothea Blostein, “Pen Acoustic Emissions
comparison. Also number of inversions in the direction of for Text and Gesture Recognition”, 10th International Conference
acceleration and velocity profiles with combination of ‘ll’ letters on Document Analysis and Recognition, 2009.
[11] Laurence Likforman-Sulem, Anna Esposito, Marcos Faundez-
were measured. The analysis was carried out based on mean
Zanuy, Stephan Clemencon and Gennaro Cordasco, “EMOTHAW:
scores of measurements for each subject. The results showed A Novel Database for Emotional State Recognition from
significant disturbance in kinetics of handwriting movements in Handwriting and Drawing”, IEEE Transaction on Human-Machine
PD subjects as compared to healthy normal people [41]. The Systems, 2017.
summary of this research study is tabulated in Table 1. [12] Essentials of Machine Learning Algorithms, Available from:
https://www.analyticsvidhya.com/blog/2017/09/common-machine-
Table 1: Summary of disease and related important writing features and learning-algorithms
clinical tests. [13] Anova Test, Available from:
Health/Mental Important Feature Clinical Test http://www.statisticshowto.com/probability-and-
Illness statistics/hypothesis-testing/anova/
Alzheimer In-air time MMSE [14] Naive Bayes Classifier, Available from:
MCI Orientation, Slant, Ordering, MODA https://en.wikipedia.org/wiki/Naive Bayes classifier
Baseline [15] K-means Algorithm, Available from:
Dysgraphia Temporal and Spatial measures HPSQ https://en.wikipedia.org/wiki/K-meansclustering
Schizophrenia Avg. Normalized Jerk, Velocity EPS [16] Random Forest Tree Generation, Available from:
http://dataaspirant.com/2017/05/22/random-forest-algorithm-
Autism In-air time and slant, Sizing, IQ
machine-learing/
Tortuosity
[17] Alzheimer’s disease, Available from:
Depression, Anxiety, In-air, On-paper time DASS, DSM-
https://simple.wikipedia.org/wiki/Alzheimer’s disease
Stress IV
[18] Perla Werner, Sara Rosenblum, Gady Bar-On, Jeremia Heinik and
Parkinson Velocity Jerk, Kinetic Measures UPDRS
Amos Korczyn, “Handwriting Process Variables Discriminating
Mild Alzheimer’s Disease and Mild Cognitive Impairment”, The
Journals of Gerontology: Series B, vol 61, Issue 4, Pages P228–
5. Conclusion P236, 2006.
[19] Paola Fontana, Francesca Dagnino, Leonardo Cocito and Maurizio
Handwriting is also called as brain writing. It is used to predict Balestrino, “Handwriting as a gauge of cognitive status: a novel
physical, emotional and mental health of an individual. In this forensic tool for posthumous evaluation of testamentary capacity”,
Neurol Sci., 29(4):257-61, 2008.
paper, a generalized system of digital handwriting analysis for
[20] Murad Badarna, Ilan Shimshoni, Gil Luria and Sara Rosenblum,
diagnosis of health and mental issues has been presented. A model “The Importance of Pen Motion Pattern Groups for Semi-
of machine leaning has been used for training and testing. Automatic Classification of Handwriting into Mental Workload
Classes”, Cognitive Computation, vol. 10(2): 215-227, 2017.
The research work carried out to diagnose diseases such as
[21] Understanding Dysgraphia By Erica Patino, Available from:
Alzheimer, Mild Cognitive Impairment, Dysgraphia, https://www.understood.org/en/learningattentionissues/childlearnin
Schizophrenia, Autism, Parkinson’s disease and mental illness g-disabilities/dysgraphia/understanding-dysgraphia
based on digital handwriting analysis has been reviewed in this [22] Jiri Mekyska, Marcos Faundez-Zanuy, “Identification and Rating of
paper. Patient’s handwriting has been obtained using WACOM’s Developmental Dysgraphia by Handwriting Analysis”, IEEE
INTUOS digitalizing tablet. The experiments and results are Transactions on Human-Machine Systems, vol. 47, Issue: 2, 2017.
summarized. It is found that features related to motion, time and [23] Sara Rosenblum, Patrice L. Weiss and Shula Parush, “Handwriting
pressure are very helpful for diagnosis of health and mental evaluation for developmental dysgraphia: Process versus product”,
Kluwer Academic Publishers, 2004.
diseases using digital handwriting analysis approach.
[24] Sara Rosenblum and Gideon Dror, “Identifying Developmental
Dysgraphia Characteristics Utilizing Hand- writing Classification
References Methods”, IEEE Transactions on Human-Machine Systems, vol.
47, Issue: 2, 2017.
[25] Schizophrenia, Available from:
[1] Cronje, Pierre E. and Hester E Roets, “Graphology in Psychological https://faculty.washington.edu/chudler/schiz.html
Assessment: A Diagnosis in Writing”, Universal Journal of [26] Michael P. Caligiuri, Hans-Leo Teulings, Charles E. Alexander B.
Psychology, vol. 1, No. 4, pages: 163-168, 2013. Niculescu, James Lohr, “Handwriting Movement Analyses for
[2] Bora Ugurlu, Rembiye Kandemir, Aydn Carus, Ercan Abay, “An Monitoring Drug-Induced Motor Side Effects in Schizophrenia
Expert System for Determining the Emotional Change on a Critical Patients Treated with Risperidone”, Hum Mov Sci., vol. 28(5):
Event Using Handwriting Features”, TEM Journal, vol. 5, Issue 4, 633–642, 2009.
Pages 480-486, 2016. [27] M. Borjkhani, M. Ahmadlou, F. Towhidkhah, “Extracting Reliable
[3] Emotions and Feelings, Available from: Handwriting Kinematic Features by Using Neural Network for
http://graphicinsight.co.za/emotions.htm. Diagnosing Schizophrenia Disease”, Biomedical Engineering
[4] Handwriting analysis and health, Available from: Conference, 2008.
http://www.handwriting-graphology.com/handwriting-analysis-and- [28] Crystal R. Blyler, Brendan A. Maher, Theo C. Manschreck and
health-graphopathology. Wayne S. Fenton, “Line drawing as a possible measure of
[5] Learn Graphology, Available from: http://www.handwriting- lateralized motor performance in schizophrenia”, Schizophrenia
graphology.com/learn-graphology/ Research, vol. 26, Issue 1, Pages 15-23, 1997.
[6] Zainab Harbiab, Yulia Hicksa and Rossitza Setchia, “Clock [29] What is Autism in Simple Terms, Available from:
Drawing Test Digit Recognition Using Static and Dynamic https://www.dealwithautism.com/whatisautisminsimpleterms/
Features”, Procedia Computer Science, vol. 112, Issue C, [30] Sara Rosenblum, “Children with high-functioning autism spectrum
September 2017, Pages 1641-1650, 2017. disorder show unique handwriting patterns”, University of Haifa.
Available from:
511 International Journal of Engineering & Technology

https://www.sciencedaily.com/releases/2016/06/160601084649.htm
.
[31] Nicci Grace, Peter Gregory Enticott, Beth Patricia Johnson, Nicole
Joan Rinehart, “Do Handwriting Difficulties Correlate with Core
Symptomology, Motor Proficiency and Attentional Behaviours?”,
Journal of autism and developmental disorders, 2017.
[32] Azadeh Kushki, Tom Chau, Evdokia Anagnostou, “Handwriting
Difficulties in Children with Autism Spectrum Disorders: A
Scoping Review”, Journal of Autism and Developmental
Disorders, vol. 41, Issue 12, pp 1706–1716, 2011.
[33] What is Depression, Anxiety and Stress, Available from:
https://www.psychologytoday.com/basics/depression
[34] Psychology Today, Available from:
https://www.psychologytoday.com/
[35] DASS Assessment, Available from:
https://openpsychometrics.org/tests/DASS/
[36] Clara Rispler, Gil Luria, Allon Kahana, Sara Rosenblum, “Mood
Impact on Automaticity of Performance: Handwriting as
Exemplar”, Cognitive Computation, 2018.
[37] Sara Rosenblum, Perla Werner, Tal Dekel, Ilya Gurevitz and
Jeremia Heinik, “Handwriting process variables among elderly
people with mild Major Depressive Disorder: A Preliminary
Study”, Aging Clinical and Experimental Research, vol. 22,
Issue 2, pp 141–147, 2010.
[38] Parkinson’s disease, Available from:
https://simple.wikipedia.org/wiki/Parkinson
[39] Michael P. Caligiuri, Hans-Leo Teulings, J. Vincent Filoteo, David
Song, James B. Lohr, “Quantitative measurement of handwriting in
the assessment of drug- induced parkinsonism”, Human Movement
Science, vol. 25, pp 510–522, 2006.
[40] Naiqian Zhi, Beverly Kris Jaeger, “Toward Monitoring
Parkinson’s through Analysis of Static Handwriting Samples”,
IEEE Journal of Biomedical and Health Informatics, vol.: 21, Issue:
2, 2017.
[41] O. Tucha, L. Mecklinger, J. Thome, A. Reiter, G. L. Alders, H.
Sartor, M. Naumann, and K. W. Lange, “Kinematic analysis of
dopaminergic effects on skilled handwriting movements in
Parkinson’s disease”, Journal of Neural Transmission, vol. 113,
Issue 5, pp 609–623, 2006.