EMG REMOVAL FROM EEG BASED ON SOBI AND PSD METHODS

Abstract

This thesis presented an efficient method of separating EMG artifacts from EEG signals containing epileptic spikes, specifically helps doctors accurately diagnose epilepsy by reduced muscular artifacts in EEG. Firstly, the noisy signals are pre-processed by a combined filter which consists of a low-pass, a high-pass and a notch filters. After that, we focused on blind source separation method - SOBI, on the basis that thesis evaluated the quality of this algorithm with simulated signals and real measured signals from patients. In the next step, the power spectrum density - PSD method is applied to identify the EMG source and take the compensation on each channel. We also investigated the effect of the frequency range to this identification. The algorithm is tested by simulated and experiment signals. The results were very good signal over noise reduction can represent clearly the epileptic spikes in the EEG may help clinicians accurately diagnose the patient's situation.