AccessBridge is an advanced AI-powered accessibility communication tool designed to bridge communication gaps for individuals with speech and hearing impairments. Leveraging state-of-the-art machine learning models, AccessBridge provides real-time speech-to-text transcription, intelligent text simplification, and personalized speech adaptation to enhance accessibility and communication independence.

To create an inclusive digital environment where communication barriers are eliminated through innovative AI technology, empowering individuals with disabilities to communicate effectively and independently.

Omer Tariq
Ph.D. Candidate in Artificial Intelligence
Korea Advanced Institute of Science and Technology (KAIST)
📧 Contact | 🔗 LinkedIn

• Multi-format Audio Support: Microphone recording, file upload (WAV, MP3, M4A)
• Wav2Vec2 Integration: Facebook's state-of-the-art speech recognition model
• Adaptive Processing: Custom user profile adaptation for improved accuracy
• Noise Robustness: Enhanced performance in various acoustic environments

• Cognitive Accessibility: BART-based text simplification for easier comprehension
• Adaptive Complexity: Adjustable simplification levels based on user needs
• Context Preservation: Maintains meaning while reducing linguistic complexity
• Real-time Processing: Instant text transformation for immediate accessibility

• Speech Adaptation Models: Machine learning-based personalization
• Pattern Recognition: K-means clustering for speech characteristic identification
• Progressive Learning: Continuous improvement through user interaction
• Profile Management: Create, save, and switch between multiple user profiles

• Tacotron2 Integration: Natural-sounding speech synthesis
• Voice Customization: Multiple voice styles and speaking rates
• Accessibility Focus: Optimized for users with hearing impairments
• Real-time Generation: Fast audio synthesis for immediate feedback

• Usage History: Detailed transcription and interaction logs
• Performance Metrics: Accuracy tracking and improvement suggestions
• Export Capabilities: CSV export for external analysis
• Visualization Tools: Audio waveform analysis and feature extraction

• Speech Recognition: Facebook Wav2Vec2-base-960h
• Text Simplification: Facebook BART-large-CNN
• Text-to-Speech: Coqui TTS Tacotron2-DDC
• Machine Learning: scikit-learn, PyTorch

• Interface: Gradio Web UI
• Audio Processing: torchaudio, librosa
• Data Science: pandas, numpy, matplotlib
• Deployment: Python 3.10+, Conda environment management

• Linux (Ubuntu 20.04+)
• macOS (Big Sur+)
• Windows 10/11

• Python 3.10 or higher
• Conda package manager
• 8GB+ RAM (16GB recommended)
• GPU support optional but recommended

# Clone the repository
git clone https://github.com/your-username/AccessBridge.git
cd AccessBridge

# Create conda environment
conda create -n accessbridge python=3.10
conda activate accessbridge

# Install dependencies
conda install -c conda-forge libstdcxx-ng libgcc-ng portaudio numpy scipy
pip install -r requirements.txt

# Launch the application
python main.py

# Create and activate environment
conda create -n accessbridge python=3.10
conda activate accessbridge

# Update conda packages
conda update --all

# Linux (Ubuntu/Debian)
sudo apt-get update
sudo apt-get install -y build-essential portaudio19-dev

# macOS
brew install portaudio

# Windows
# Use conda-forge packages (recommended)

# Core ML libraries
pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu
pip install transformers datasets

# Audio processing
pip install librosa soundfile

# Web interface
pip install gradio

# Additional dependencies
pip install scikit-learn pandas matplotlib joblib
pip install TTS

Models are automatically downloaded on first run. Ensure stable internet connection for initial setup (~2GB download).

torch>=2.0.0
torchaudio>=2.0.0
transformers>=4.30.0
gradio>=3.35.0
librosa>=0.10.0
scikit-learn>=1.3.0
pandas>=2.0.0
matplotlib>=3.7.0
numpy>=1.24.0
joblib>=1.3.0
TTS>=0.15.0

1. Launch Application

   python main.py

2. Access Interface

   • Open browser: http://localhost:7860
   • Navigate through intuitive tabs

3. Speech Transcription

   • Record audio or upload file
   • Select user profile
   • Enable text simplification if needed
   • Click "Transcribe Audio"

# Profile creation workflow
1. Navigate to "User Profiles" tab
2. Enter profile name
3. Record multiple speech samples
4. Provide transcriptions
5. Create adaptive model

• Automatic: Enable checkbox for instant simplification
• Manual: Use dedicated Text-to-Speech tab
• Customizable: Adjust complexity levels in settings

• History Export: Download transcription logs as CSV
• Performance Tracking: Monitor accuracy improvements
• Usage Analytics: Analyze interaction patterns

• 🌐 Multilingual support
• 🎨 UI/UX improvements
• 🧠 Advanced ML models
• 📱 Mobile optimization
• ♿ Enhanced accessibility features

• Real-time streaming transcription
• Multiple language support
• Advanced user analytics
• Mobile application

• Cloud deployment options
• API integration
• Advanced personalization
• Collaborative features

This project is licensed under the MIT License