Interactive Demo
Voice Analysis Demo
Record or upload a sustained vowel sound and send it to the trained CNN-LSTM model for real-time Parkinson's disease voice analysis.
Research Demo — Not a Clinical Tool
This tool is for educational research purposes only and is not a medical diagnostic tool. Results reflect pattern similarity to the training dataset and do not constitute medical advice.
Drag & drop or click to upload
Supported: .wav (recommended), .mp3, .m4a
See what a good recording looks like before you start.
Demo Settings
Show Spectrogram Details
Reveals the backend-generated mel-spectrogram and pipeline steps
Analysis Pipeline
Upload or record a voice sample to begin
Classification Results
Parkinson's Pattern
—Healthy Pattern
—Results appear after analysis
About This Demo
- Audio is sent to a local FastAPI backend running on port 8000
- The backend preprocesses audio and runs a trained CNN-LSTM model
- Model trained on the PC-GITA dataset with 5-fold nested cross-validation
- Waveform decoding uses the browser Web Audio API
- Toggle spectrogram details to see the actual mel-spectrogram from the backend