Overview
Designed and optimized Python-based data pipelines on AWS for high-frequency EEG signals, applying wavelet and Hilbert transforms with sliding-window analysis to extract time-series features for cognitive state modeling.
Problem
EEG signals are high-dimensional, noisy, and require domain-specific feature extraction before any ML model can be applied. Raw signals from research subjects needed to be processed in near-real-time for downstream cognitive state classification.
Approach
- Wavelet transforms for time-frequency feature extraction
- Hilbert transforms for instantaneous amplitude and phase
- Sliding-window analysis for temporal feature engineering
- AWS pipeline for scalable data management
- Regression models correlating multi-modal physiological data
Results
| Metric | Value | |--------|-------| | R² (regression) | up to 0.83 | | Signal type | High-frequency EEG | | Platform | AWS |
Tech Stack
Python AWS MATLAB scikit-learn Wavelet Signal Processing