Brain-computer interface design based on wavelet packet transform and SVM

For the BCI research to classify the different imagined movements of both left and right hands, a method using wavelet packet decomposition for feature extraction and using SVM for pattern classification was adopted. Firstly discusses the wavelet packet transform in depth and brings out an idea of taking wavelet packet coefficients' variance as feature into account, then extracts the feature serials after wavelet packet decomposition for channel C3 and C4, finally, classify the patterns by using linear SVM. The result shows that the maximum classification accuracy is 86.43% and the feature of variance is suitable. So, the method this paper used for feature extraction and pattern classification is more efficient and simpler, and it gives a new reference for BCI.