2nd Place Winner – AIMD Spring 2025 Presentation Night
A voice-based machine learning system that detects early signs of Parkinson’s Disease by analyzing vocal biomarkers with over 85% accuracy. Built as part of a semester-long project at UTD, this tool leverages advanced signal processing and machine learning techniques to aid early diagnosis.
Parkinson's Detect uses patients' voice recordings to extract key vocal features such as:
- Pitch variations
- Jitter
- Shimmer
- Tremor indices
These features are then processed through a robust ML pipeline using:
- ✅ Random Forest
- ✅ Support Vector Machine (SVM)
Our system provides a non-invasive, accessible way to screen for Parkinson’s symptoms — with potential real-world impact in clinical pre-screening tools.
- Python
- Librosa – Audio signal processing
- NumPy / Pandas / Scikit-learn – Feature extraction and ML
- Matplotlib / Seaborn – Visualizations
- Streamlit – User-friendly diagnostic interface
- Upload and process voice recordings
- Extract medically relevant vocal biomarkers
- Predict Parkinson's likelihood with ML models
- Visualize feature contributions and diagnosis confidence
Through this project, I gained deep hands-on experience in:
- Feature engineering for biomedical signals
- ML model selection, training, and hyperparameter tuning
- Audio preprocessing and real-time voice analysis
- Building ML interfaces with Streamlit
This was a collaborative effort by our AIMD Spring 2025 team - Romita Veeramallu, Krithika Muthukumar, Shariq Hassan, Pavan Arani, Rajat Singh, and myself. I'm grateful to my teammates and mentors for their insights, support, and innovation throughout this project.