EEG Signal Classification by Combining Discrete Wavelet Transform and Empirical Mode Decomposition
The most complicated body part is the human brain which can regulate and perform many activities and can perceive. A brain that is working in good condition is capable of producing responses to the received signal and then forward the information to the entire body. A few syndromes of the brain adversely affect the interactions between the brain and the remaining part of the body and thereby limiting the transmission of signals. The device that is capable of helping diseased people with neural abnormalities is termed Brain-Computer Interfaces (BCIs) which assist in evaluating the activities of the brain, interpreting and categorizing different actions of the brain, and later forwarding the signals to the external system and possibly back to the body. One of the significant aspects of optimizing the efficiency of the brain-computer interface is the rate of recognition for the categorization of neural activity. Therefore, the technique of extracting feature is presented in this work which is based on the entropy approach, discrete wavelet transform (DWT), and empirical mode decomposition (EMD). By employing various characteristics like standard deviation, mean, RMS, entropy, and average, the feasibility of the designed framework can be analyzed. Initially, a series of low-frequency signals along with DWT is obtained from the decomposition of electroencephalogram (EEG) signal, further with the help of EMD the sub-band signal is processed to obtain the number of which stationary time series known as intrinsic mode functions (IMFs). Meanwhile, for the reconstruction of the signal, suitable IMFs are chosen.decomposed into a series of narrowband signals with DWT, then the sub-band signal is decomposed with EMD to get a set of stationary time series, which are called intrinsic mode functions (IMFs). Meanwhile, for the reconstruction of the signal, the suitable IMFs arechosen.Therefore, the associated feature vector is extracted with the help of the computed entropy of the received signal. At last, classification is carried out through k-nearest neighbors (KNN) and support vector machine (SVM). Thus, the coverage of large frequency issues during EMD is overcome by the proposed framework and thereby enhancing the classification efficiency of EEG signal motion imaging. The experimental results show that the KNN classifier is more efficient and achieves 93.84% accuracy when compared to the SVM classifier which achieves 87.64% accuracy.