In order to feed the time series features into the presentedapproach, the raw signal was split into 20 segments to constructthe sequence features to transfer each sample into a matrix with ashape (20,7). Secondly, the ten features from Table 2 are extractedfrom raw signal. After that, for each time segment there are 70feature values within a feature vector. Ultimately, the shape of anintact feature matrix is (20, 70).