Using Adaptive Surface EMG Envelope Extraction for Onset Detection: A Preliminary Study on Upper Limb Amputees

Surface electromyography is a valid and widely used tool for the characterization and the intention of human movement in healthy and pathological subjects; in particular, the thorough identification of the transient phase of EMG activity is an important problem to be solved for both myoelectric control applications for prosthetics and rehabilitation and in movement analysis in general. Among the possible algorithmic solutions, the ones based on the statistical properties of the signal have been considered able to yield stable performance in a variety of different scenarios. In this paper, an adaptive and statistically optimised algorithm for the extraction of the amplitude envelope is exploited for the onset detection from EMG data coming from the shoulder muscles of two upper limb amputees. In particular, onset events have been detected from the optimised point-by-point window length of the adaptive filter, as the instants in which this time series reaches a local minimum, and compared with those coming from visual inspection of accelerometer data from the shoulder. These preliminary results show how using such techniques can yield acceptable performance, supporting the hypothesis of exploiting such an algorithm for the improvement of the performance of myoelectric control algorithms to be applied in clinical contexts.