Removing Jaw Clench, Teeth Squeeze and Forehead Movement EMG Artifacts from EEG Signal using Dynamic Size Segmentation and Multilevel Decomposed Wavelet with Adaptive Thresholding

Background: Jaw clench, teeth squeeze and forehead movement EMG artifacts affected the Electroencephalogram and suppress clinically important information regarding brain neural activity. EMG artifact contaminates the EEG signal and creates an obstruction in proper diagnosis of patients suffering from brain related diseases. EEG recording takes place almost from 30 to 40 minutes and it is often that EMG artifacts get embedded in important brain neural activity. Hence it is required to suppress the EMG artifacts. Materials and Methods: 16 channel EEG signals are acquired with EMG artifacts. The subject is instructed to do jaw clenching, forehead movement, teeth squeezing at different instances during the time of recording. Sampling rate chosen is 1024 Hz. The present work tried to remove these artifacts using dynamic size segmentation of EEG signal and multilevel decomposed wavelet enhanced independent components. This new method not only removes the artifact but also estimates data present in the time span of artifact region. In this present work, three methods of artifact removal are discussed and compared. The first method is wavelet Enhanced Independent Component analysis with static segmentation; the second method is wavelet Enhanced independent components with dynamic size segmentation of EEG signal and the third method is multilevel decomposed wavelet with adaptive thresholding with the time--frequency domain approach. The statistical parameters like PSNR, PSD, RMSE, standard deviation are used for performance measurements. Results: Proposed method in study, namely automatic dynamic segmentation with adaptive thresholding shows superior performance than other two methods discussed. It suppresses EMG artifacts significantly. Conclusion: Extensive lab results showed that dynamic size segmentation is a better tool to remove out EEG artifact over static size segmentation, but the third method is most suitable for estimation of data present in the time span of artifacts.