Automatic Signal Segmentation using the Fractal Dimension and Weighted Moving Average Filter

In many applications of the signal processing such as automatic analysis of EEG signal, it is needed that signal is split to smaller parts that each part has the same statistical characterizations such as the amplitude and frequency. This act has been called signal segmentation. In this paper, the signal is initially filtered by weighted moving average (WMA). Not only WMA can emphasize recent events which this act is very important in the signal segmentation, but it can also detect important underlying unadulterated from of the time series by attenuating its short-term variations. After filtering the signal, fractal dimension (FD) of the signal is computed and used as a feature for automatic segmentation of the signal. The proposed method has been applied on the synthetic and real EEG signals and then, this method has been compared with improved nonlinear energy operator (INLEO) method which is known as a good method for segmenting a signal. The simulation results indicate that these proposed techniques have greater accuracy compared with previous methods.