Figure 4.

Windowing scheme of sEMG data in the continuous feature extraction. The size of each epoch of individual shapes is T. The size of sliding window for detecting threshold crossing is first set as l with a window increment of s. When the RMS value of the sliding window crossing the threshold lasts for B segments, the onset of the window will be the epoch onset of individual shapes. Next, the size of successive analysis windows for classification will change to L with a window increment of S. RMS, MAV, and aEMG are calculated at each L intervals. The sEMG data for classification is divided into C segments for every L that is the length of integrated samples as a feature extraction and the start point is shifted every S. Although six channels of sEMG data are used, only the main channel is shown here for illustrative purposes.