Decoding Movement Initiation in a Brain-computer Interface for Real-Time Speech Synthesis

In this study, we conducted a time frequency analysis of EEG signals collected in a limb motor imagery brain-computer interface (BCI) paradigm to identify features useful for predicting and decoding onset and termination of imagined movement. Participants used the BCI to control a formant frequency speech synthesizer to produce three different vowel sounds. A previous analysis focused on decoding formant frequencies for BCI control. The results of this study found increases in theta-band amplitude (3–7 Hz) across all participants at the beginning and ending of motor imagery periods (approximately 500 ms after the onset cue and 500 ms after the termination cue). A Kalman Filter was then designed to decode the onset and termination signals from a calibration period and tested on the experimental EEG data recorded from each participant. Initial results show the Kalman filter predicted onset and termination from unseen data that correlates with increases in theta-band amplitude. Our findings, when combined with our prior formant frequency decoding, will improve the overall performance of the BCI by focusing the decoder only on time intervals during which participants are actively engaged in the BCI task.