A novel dot-plot algorithm for surface EMG signal segment identification

Abstract

Segmentation of surface EMG signal often involve in many scenarios including motion classification in robotic prostheses and motion segment identification. Many of them require a threshold that is predefined or a training data set. The objective of this paper is to find a way to perform segmentation without a threshold or training data set. Dot plot analysis has been widely used in bioinformatics to identify similar segments between proteins or DNA. The philosophy behind dot plot analysis can be applied to perform surface EMG signal segmentation. The properties of the new algorithm is examined. The major advantage of the dot-plot segmentation algorithm is that threshold is no long need to be estimated, instead the minimal length of a segment in a time series signal need to be declared.