Abstract
Conference Title: 2016 24th European Signal Processing Conference (EUSIPCO) Conference Start Date: 2016, Aug. 29 Conference End Date: 2016, Sept. 2 Conference Location: Budapest, Hungary A new efficient and user-independent technique for the detection of muscle activation (MA) intervals is proposed based on Gaussian Mixture Model (GMM) and Ant Colony Classifier (AntCC). First, time and frequency features are extracted from the surface electromyography (sEMG) signals. Then, GMM is used to cluster these extracted features into burst & non burst. Those features with their class name are then used as the input for the AntCC algorithm in order to derive classification rules. Finally, the obtained rules are used to detect sEMG activation timing of human skeletal muscles during movement. The performance of the proposed technique is demonstrated by means of synthetic simulated sEMG signals and real ones. The proposed technique is then compared to two previously published techniques: wavelet transform-based method [1] & double threshold-based method [2]. It is concluded that our technique outperforms those methods and significantly improves the accuracy of good MA timing detection. Moreover, to our knowledge, the proposed technique is the first user-independent one since no tuning parameters are required. Our findings show that the proposed method is convenient for automatically processing large amounts of sEMG signals with performance beyond that of the state of the-art methods.