A Novel Signal Segmentation Method based on Standard Deviation and Variable Threshold

Decomposition of non-stationary signals such as electroencephalogram (EEG) and electrocardiogram (ECG) into stationary or quasi-stationary, signal segmentation, is a wellknown problem in many signal processing applications. Previous methods for segmenting a signal had problems such as slow speed, low performance, and several parameters which must be defined experimentally. In this paper a new method based on standard deviation and variable threshold has been suggested. The standard deviation can indicate changes in the amplitude and/or frequency that it is the purpose of the signal segmentation. Since the standard deviation isn’t able to indicate the effect of the shift in a signal, the proposed method utilizes the integral as a pre-processing level. Also, to improve the efficiency of the proposed method we use variable threshold. In order to evaluate the performance of this method, we use synthetic and real EEG signals. In EEG signals to remove destructive noises like EMG and EOG, we propose to use discrete wavelet transform (DWT). The obtained results indicate superiority of the proposed method in signal segmentation.