| Abstract [eng] |
Surface elecromyography is a common and effective method for muscle activity registration. With sEMG electrodes signals of multiply motor units are registered. At the same time, signal is induced with noise and artifacts, which are associated with movement and partial comming off of electrodes, subcutaneous fat tissue movement and electrical activity of the heart. The aim of this work is to separate movement artifacts as precisely as possible, by checking every artifact with cross-correlation method. Savitzky-Golay and moving average filters were compared for filtering the correlation curve. It was determined, that Savitzky-Golay filter with 55 ms window is the most effective filter for this method, because it removes unwanted artifacts form correlation curve with least possibility to lose the important artifacts in muscle activity periods. Another problem of electromyograpy signals is manual signal segmentation, which demands a lot of time and does not ensure that the muscle activity sections will be segmented precisely, without losing the end and the beginning of the activity segment, and without adding parts of the signal with no activity. These problems may affect the frequency parameters of the surface electromyohraphy signals. For signal segmentation to activity and resting sections, two methods are used: 1) segmentation by amplitude and 2) segmentation by double standard deviation. To sum up, when performing the segmentation of activity and resting sections of sEMG signal, the activity recognition error if bigger when segmentating by amplitude, that segmentation by double standard deviation. |
