EMOVOCAL

Voice-Based ADHD Detection Using Machine Learning

An advanced web application leveraging voice analysis and machine learning for ADHD characteristic detection

📋 Table of Contents

• Overview
• Features
• Technology Stack
• Project Structure
• Prerequisites
• Installation
• Usage
• API Documentation
• Development
• Deployment
• Security
• Contributing
• License

🔍 Overview

EMOVOCAL is a sophisticated Flask-based web application that performs real-time voice analysis for ADHD detection. The system processes audio recordings through state-of-the-art machine learning models trained on eGeMAPs (extended Geneva Minimalistic Acoustic Parameter Set) features to identify potential ADHD indicators in speech patterns.

✨ Features

• Real-time Voice Analysis

  • Instant processing and feedback
  • Support for MP3 and WAV audio formats
  • Dynamic progress tracking

• Advanced Audio Processing

  • Automated audio file segmentation
  • High-precision feature extraction using eGeMAPs
  • Multi-threaded processing for optimal performance

• Machine Learning Integration

  • State-of-the-art ADHD detection models
  • Probability-based classification
  • Continuous model improvement capabilities

• User Interface

  • Modern, responsive design
  • Intuitive user experience
  • Real-time progress visualization
  • Detailed results presentation

🛠 Technology Stack

• Backend

  • Python 3.13.2
  • Flask 3.1.0
  • OpenSMILE 2.5.1
  • scikit-learn 1.6.1

• Audio Processing

  • librosa 0.11.0
  • soundfile 0.13.1

• Data Analysis

  • pandas 2.2.3
  • numpy 2.1.3
  • matplotlib 3.10.1
  • seaborn 0.13.2

• Development Tools

  • joblib 1.3.2
  • tqdm 4.67.1

📁 Project Structure

.
├── app.py                 # Main Flask application
├── create_predict_data.py # Audio processing pipeline
├── predict.py            # ML model inference
├── static/              # Static assets
│   ├── css/            # Stylesheets
│   ├── js/             # JavaScript files
│   └── img/            # Images and icons
├── templates/          # HTML templates
│   ├── index.html     # Main interface
│   └── results.html   # Results display
├── uploads/           # Temporary file storage
└── requirements.txt   # Dependencies

📋 Prerequisites

• Python 3.13.2 or higher
• FFmpeg (for audio processing)
• gcc (for building dependencies)
• 4GB RAM minimum
• 2GB free disk space

🚀 Installation

1. Clone the repository:

git clone https://github.com/Phoenix-ryan-1111/ADHD_classification.git
cd EMOVOCAL

2. Create and activate virtual environment:

python -m venv venv
# Windows
venv\Scripts\activate
# Unix/MacOS
source venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

💻 Usage

1. Start the application:

python app.py

2. Access the web interface:

https://127.0.0.1:5000

Or

https://truthful-miracle-production.up.railway.app/

3. Upload an audio file (MP3/WAV)
4. Wait for analysis completion
5. Review detailed results

🔌 API Documentation

Endpoints

• GET / - Main application interface
• POST /upload_file - Audio file upload endpoint
  • Accepts: multipart/form-data
  • Returns: Server-Sent Events (SSE) with analysis progress

Response Format

{
  "success": true,
  "prediction": "ADHD/Non-ADHD",
  "probability": "float (0-1)",
  "percentage": "float (0-100)"
}

🛠 Development

The application implements:

• Server-Sent Events for real-time updates
• Responsive design principles
• Modern CSS with flexbox/grid
• Event-driven JavaScript architecture

🚀 Deployment

Configured for Railway deployment with:

• Python 3.13.2 runtime
• Nixpacks builder
• Automatic environment configuration
• Production-grade server settings

🔒 Security Features

• Secure file upload handling
• File size restrictions (1000MB max)
• File type validation
• Secure filename processing
• Automated temporary file cleanup
• Production environment detection

🤝 Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Made with ❤️ by the EMOVOCAL Team