Analysis of surface electromyography signal features on osteomyoplastic transtibial amputees for pattern recognition control architectures. (February 2018)

• Record Type: Journal Article
• Authors: Garikayi, Talon; Van den Heever, Dawie; Matope, Stephen
• Is Part Of: Biomedical signal processing and control. Volume 40(2018)
• Page Start: 10
• Page End: 22
• Publication Date: 2018-02
• Subjects: Myoelectric control -- SENIAM -- Electromyography -- Pattern recognition -- Prosthetic -- Amputee; 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.2017.09.007
• Languages: English
• ISSNs: 1746-8094

Abstract

Graphical abstract: Highlights: Residual amputated lower limb generates electromyography signals for myoelectric control. There is significant difference on electromyography signal features between amputee and normal subject electromyography signal. There is no significant difference on electromyography signal features among osteomyoplastic amputees. Electromyography signals from amputated lower limbs possess low signal strength for both time domain and frequency domain features. Weak signals from an amputated lower limb can be used for classification. Abstract: This paper presents the characterisation of electromyography signals for the purpose of controlling a powered prosthetic ankle using pattern recognition algorithms. The goal is to identify the specific muscles that can be used to guarantee optimal control of a multichannel powered prosthetic ankle. SENIAM and ISEK protocols were used for signal acquisition, processing and reporting. A set of paired surface electrodes were placed above selected muscles on the residual limb. Participants were instructed to perform normal gait. The signals were recorded, labelled and analysed using the Vicon Nexus Motion Capturing System and Noraxon Myomotion System. Signal processing was performed using MR3 Software and further post processing was performed using Matlab. Time and frequency domain features were analysed. The protocol revealed that the tibialis anterior, medial and lateral gastrocnemius muscles actively generate myoelectric signals on the residual limb. A total of 12 time domain and 4 frequency domain features were successfully extracted and used in the analysis. The tibialis anterior muscle was identified as a candidate for classifying dorsiflexion with a mean amplitude of 35.08 μV. The soleus muscle was inaccessible on the amputated leg and as a result only the medial and lateralis gastrocnemius muscles, with 17.40% signal power and 43.73% mean amplitude as compared to the soleus, were available for plantarflexion. There was significant difference (p < 0.05) between features from the amputated residual limb and those from the intact normal leg. However, there was no significant difference (p > 0.05) between signal features from two different participants. Sagittal plane movements were linearly discriminated with 100% accuracy for tibialis anterior and medial gastrocnemius. However, lateralis gastrocnemius exhibited a 0.0769% classification error as a result of the amputation technique.