Comparative study of simple feature extraction for single-channel EEG based classification

In this paper, the single-channel EEG based classification systems using simple extracted features are investigated. Each classification system contains the following stages: data acquisition, signal decomposition, feature extraction, and classification. In addition to using the filter bank and empirical mode decomposition (EMD) methods for signal decomposition, a sparse discrete wavelet packet transform (DWPT) is proposed. The filter bank based classifiers provide the best classification accuracy while the EMD and DWPT classifiers offer acceptable classification accuracy. As a shortcoming, the EMD based classifiers demand the large computation load for feature extraction. However, the sparse DWPT based classifiers have less computation requirement. The usefulness and efficiency of the proposed simple features are validated.