Novel algorithm for conventional myocontrol of upper limbs prosthetics. (March 2020)

• Record Type: Journal Article
• Authors: Benchabane, S.I.; Saadia, N.; Ramdane-Cherif, A.
• Is Part Of: Biomedical signal processing and control. Volume 57(2020)
• Publication Date: 2020-03
• Subjects: Time domain myo-control -- Artificial hands -- Bionics hands -- Multi channel EMG recoding -- Humanoid prosthesis -- Myoelectric interface -- Non-invasive electromyography -- Upper limb prosthetics -- Pattern recognition; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101791
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The signal magnitude describing the myoelectric signal at a given electrode location will be repeatable for a given state of muscle activation and different between activation states. Adaptive thresholds were computed according to sEMG trend in time domain. Functions' thresholds are updated according to Difference Absolute Standard Deviation Value (DASDV) and to Mean Absolute Value (MAV). The algorithm have been assessed regarding to muscle fatigue and donning/doffing. Abstract: 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.