Wearable & Real-Time Physiological Signal Processing
Robust, real-time ML for noisy wearable and ambulatory biosignals.
I develop end-to-end pipelines that make physiological signals usable in real-world, on-body conditions.
- Real-time movement-artifact removal for ambulatory EEG using constrained ICA with Lagrange multipliers, improving signal-to-noise ratio and online classification accuracy (IEEE TNSRE); released as the open-source cIOL toolkit.
- MobileBCI, an around-ear EEG dataset recorded while standing, walking, and running, with end-to-end preprocessing and analysis code (Scientific Data).
- EMG reference channels to remove muscle artifacts from EEG during acquisition across multiple datasets.