Classification of electrocorticography based motor imagery movements using continuous wavelet transform

Recently electrocorticography (ECoG) has emerged as a potential tool for Brain Computer Interfacing applications. In this paper, a continuous wavelet transform (CWT) based method is proposed for classifying ECoG motor imagery signals corresponding to left pinky and tongue movement. The total experiment is carried out with the publicly available benchmark BCI competition III, data set I. The L₂ norms of the CWT coefficients obtained from ECoG signals are shown to be separable for the two classes of motor imagery signals. Then the L₂ norm based features are subjected to principal component analysis, yielding a feature set with lower dimension. Among various types of classifiers used, support vector machine based classifiers have been shown to provide a good accuracy of 92% which is shown to be better than several existing techniques. In addition, unlike most of the existing methods, our proposed method involves no pre-processing and thus can have better potential for practical implementation while requiring much lower computational time in extracting the features.