Voice-Based Cognitive Decline Detection

This project implements a proof-of-concept pipeline for detecting cognitive decline markers in speech samples. It uses both acoustic and linguistic features to analyze speech patterns and identify potential indicators of cognitive impairment.

Features

• Acoustic feature extraction (MFCC, pitch, spectral features)
• Linguistic feature extraction (word count, sentence structure, etc.)
• Unsupervised learning for pattern detection
• Risk score calculation
• Visualization of results
• Comprehensive reporting

Installation

1. Clone the repository:

git clone <repository-url>
cd memotag_assignment

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

Usage

1. Place audio samples in the audio_samples/original directory (supported formats: .wav, .mp3, .flac)

2. Run the pipeline:

python run_pipeline.py

3. Check the results:

• Risk assessment report will be displayed in the console
• Visualizations will be saved as PNG files

Project Structure

• cognitive_detection.py: Main implementation of the cognitive decline detection pipeline
• run_pipeline.py: Script to run the complete analysis pipeline
• requirements.txt: Python dependencies
• audio_samples/: Directory for audio files
• cache/: Directory for cached features and intermediate results

Requirements

• Python 3.8+
• Required Python packages are listed in requirements.txt
• Audio files in .wav, .mp3, or .flac format

Results

The pipeline generates:

• Feature importance analysis
• Risk scores for each audio sample
• Visualizations of patterns and clusters
• Detailed analysis report

Recommendations

For clinical deployment:

1. Expand dataset with validated clinical samples
2. Incorporate supervised learning with confirmed diagnoses
3. Add contextual task performance features
4. Develop longitudinal tracking
5. Validate against established assessment tools