Optimal Local Temporal Common Spatial Patterns for Single-trial EEG Classification

The classification of brain-wave such as an electroencephalogram (EEG) has an important character in brain-computer interface. Local temporal common spatial patterns (LTCSP) is a technique to capture the local temporal information of EEG signal, and has two free parameters including the number of nearest neighbors and the kernel variance, which have to be specified manually. In this paper we propose a novel method for selecting the optimal parameter for LTCSP automatically. Comparing computational time and classification rate on BCI Competition 2003 data set, it shows that our proposed method is more efficient than CSP and LTCSP.