Tracking the Dynamic Brain: Toward Real-World Brain-Computer Interfaces

Abstract

The advancement of biosensors and wearable technologies enables noninvasive, real-time measurement of human brain activity, posting potential applications in clinical diagnosis, health monitoring, and brain-computer interfaces (BCI). However, an urgent need remains for computational tools to effectively decode the brain activity, especially from unlabeled data, and to minimize the signal degradation due to noises in real-world environments. ? ? In this talk, I will present a data-driven approach, adaptive mixture independent component analysis (AMICA), for decoding brain-state changes using unlabeled electroencephalography (EEG). I will share results applying AMICA to characterize sleep stages, reveal sub-second transitions between alert and drowsy states, and uncover mental state changes during an emotion-imagery experiment. Next, I will present a software tool for automatic, real-time noise reduction for EEG-based BCIs. Finally, I will share some of the ongoing projects in real-world BCI applications.