Temporal-spatial filter algorithm for MRP feature extraction

For motor related brain-computer interface (BCI), discriminative spatial pattern (DSP) is one of the most successful algorithms in extracting the feature of movement related potential (MRP). However, by sliding window and cross validation, DSP can not catch MRP temporal feature effectively. In this paper, we propose a new algorithm called temporal-spatial filter (TSF). Given MRP signals, TSF employs alternate optimization to seek a pair of temporal-spatial projections that simultaneously maximizes the between-class scatter and minimizes the within-class scatter. The experimental results show that TSF outperforms DSP in extracting MRP feature. If combining the event-related desynchronization (ERD) and MRP, TSPF still obtains better performance than DSP.