PD Voice Research
AI Voice Analysis

AI Voice Pattern Analysis

Upload or record a sustained vowel and see how a deep learning model compares your voice against patterns learned from its training dataset.

1. Listen to an example
2. Upload or record
3. Analyze your voice
4. Review your results
1

Listen to Example RecordingsRecommended

Hear the difference between a clear recording and a poor one before you begin.

🎧

Example Recordings

Recommended

Compare a clear recording with a poor-quality recording so you know what to submit.

Good Recording Example

A clear, steady “Ahhh” recorded for at least five seconds in a quiet environment.

Poor Recording Example

An example with issues such as background noise, silence, changing volume, or an unstable vowel.

2

Upload or Record Your Voice

Use an existing audio file or record directly from your microphone.

Before You Record

  • Say "Ahhh" continuously for at least 5 seconds.
  • Record in a quiet room.
  • Speak at a normal volume.
  • Avoid background noise.
  • Hold a steady voice instead of changing pitch.

Drag & drop or click to upload

Supported: WAV · MP3 · M4A · WebM · OGG

3

Analyze Your Voice

The model converts your voice into a mel spectrogram and compares it against patterns it learned during training.

Visualization

Show the mel-spectrogram and processing steps

Voice Analysis

Add a voice recording in Step 2 to begin.

4

Your Analysis

Your voice analysis will appear here after processing.

Your voice analysis will appear here.

Complete the steps above to see your analysis.

How It Works

  • Extracts a 5-second sustained vowel segment from your recording.
  • Generates a mel-spectrogram representing your voice characteristics.
  • Runs a CNN-LSTM deep learning model trained on voice recordings.
  • Compares your voice profile against patterns learned from the training dataset.
  • Returns an AI-generated voice pattern similarity assessment.
Educational Research Notice — This application demonstrates a deep learning model developed for Parkinson's disease voice research. Results describe similarity to patterns learned from the research dataset and should not be interpreted as a medical assessment or clinical recommendation.