One of the most common issue in surface electromyography (sEMG) based myocontrol is to set a recurrent feature which allows to ensure a reliable multi degree of freedom (MDoF) prosthetic drive, mainly due to non-stationary behavior of signal. According to studies, electrode placement and shifts, variation in muscle contraction effort and muscle fatigue are the most common disturbance sources in sEMG recording, which traduces into a cumbersome donning and doffing recalibration. Many relevant works showed promising results in sEMG intent decoding; nevertheless, very few of them have addressed performance reliability over days. This research proposes a method for setting and auto-updating functions’ thresholds in non-pattern recognition time domain (TD) myocontrol and develop a simple, and wearable myoelectric interface (MEI) able to decode user’s intent. This work is expected to provide an acceptable MEI accuracy as well as resiliency to common disturbances over daily use. The testes were performed on 6 healthy users, the system accuracy (SA) obtained was 89.6 % in flexion for the five fingers, average SA deteriorates from 4.8 % over 7 days of electrodes replacement experimentation when muscle fatigue dropped SA from 2.4 %, approach relies on sEMG magnitude vectors recorded from 5 channels for computing functions’ thresholds and on difference absolute standard deviation value (DASDV) feature as updating condition. The algorithm is light computing and, provided some interesting results regarding outputs reliability. The system was produced with 35 Euros.
Read full abstract