A framework of common spatial patterns based on support vector decomposition machine

In the study of brain-computer interfaces (BCI), the techniques of feature extraction and classification play an important role, especially in classifying single-trial electroencephalogram (EEG). In previous research, many researchers have solved these problems in two separate phases: firstly use techniques such as singular value decomposition and common spatial pattern to extract features; then design classification such as linear discriminant analysis or support vector machine. In this paper, we show a framework that combines common spatial patterns (CSP) and support vector machine (SVM) to analyze single-trial EEG/ECoG dataset. We demonstrated experimental results in the data set I of ldquoBCI Competition 2005rdquo analysis with this method which gets a high level of classification accuracy on the test set.