Decoding finger movements from ECoG signals using Empirical Mode Decomposition

ECoG promises exact localization of brain sources by providing high spatial resolution and good signal quality, thus makes it the premier choice for future BCI applications. Unfortunately decoding these signals is not as straightforward as one would expect. In this work we applied a time-frequency analysis based on Empirical Mode decomposition (EMD) and Adaptive Filtering (AF) to decode and estimate the finger movement using 10 minutes-long, multi-channel ECoG signals. The dataset was recorded from three epileptic patients at Harborview Hospital in Seattle, Washington for Brain Computer Interface (BCI) Competition IV. Our proposed method showed the average correlation of 0.55 between real and predicted movement across the subjects and across fingers.