CLASSIFICATION OF EMG SIGNALS BASED ON CURVELET TRANSFORM AND RANDOM FOREST TREE METHOD

摘要

Electromyography (EMG) signal is most powerful signal processing tools for electrical activity of neuromuscular associated with a corresponding muscle. In this paper, the analysis of EMG signals using curvelet transform and Random forest tree is presented. The EMG signal including noise though dissimilar media. The curvelet transform is used for clear away noise from the surface electromyography and superior order of statistics is used for analyzing the signal. The first level is to evaluate the surface of EMG signal and extract features using curvelet transform. The second level is best EMG quality segment was chosen and the rebuilding of the useful data signal was finished using random forest classifier. The intention of this work is introducing a novel approach for discover, analyzing and classifying of EMG signals. The proposed method is applied using clinical dataset and the parameters like mean root mean square, correlation coefficient and absolute value are calculated and to get better quality of class separability. A comparison is made with other traditional methods and the EMG characteristics extracted from rebuilding of EMG signals provide the enhancement of class separability in feature space than. Statistical results shows maximum classification accuracy of 99% and higher information transfer rate is achieved.