Combination of discrete wavelet packet transform with detrended fluctuation analysis using customized mother wavelet with the aim of an imagery-motor control interface for an exoskeleton

One critical issue in brain computer interface (BCI) studies is to extract imaginary movement patterns from electroencephalograph (EEG). In this study, two different techniques -detrended fluctuation analysis (DFA) and discrete wavelet packet transform- are combined (DWPT-DFA) for feature extraction. Both approaches are known as self-similarity quantifier techniques. In wavelet technique, mother wavelets play an important role. Herein, A customized mother wavelet utilizing event related desynchronization (ERD) potential patterns are extracted and updated automatically for individual subjects. Also, three predefined mother wavelets are used, and the results are compared with the customized mother wavelet. The predefined mother wavelets are db4, db8 and coiflet 4. The soft margin support vector machine with the generalized radial basis function (SSVM-GRBF) is employed to classify the DWPT-DFA features. For the efficiency of the method, nine subjects have participated to record EEG based on the imaginary hand movements. The ERDs and features are extracted from FC1 and CP6 channels. Results show that the combination of the DWPT and DFA with the personalize ERD mother wavelet gives the best accuracy of 85.33% with p <0.001. Based on the results, we conclude that the DWPT-DFA method using the ERD mother wavelets improves significantly the efficiency of the SSVM-GRBF classifier.