Wrist and finger motion recognition via M-mode ultrasound signal: A feasibility study. (January 2022)

• Record Type: Journal Article
• Authors: Li, Jianmin; Zhu, Kun; Pan, Lizhi
• Is Part Of: Biomedical signal processing and control. Volume 71(2022)Part A
• Publication Date: 2022-01
• Subjects: M-mode ultrasound -- B-mode ultrasound -- Pattern recognition -- Human–machine interface; 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.2021.103112
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A new signal source for the wrist and finger motion recognition was investigated. M-mode ultrasound performed well for the wrist and finger motion recognition. M-mode ultrasound showed potential for the HMI applications. Abstract: With the ability to precisely detect muscle deformation, ultrasound sensing has been widely employed as a promising technique to interpret movement intentions in the field of human–machine interface (HMI). Compared to B-mode and A-mode ultrasound which had been systematically investigated, M-mode ultrasound as a kind of ultrasound had never been studied. As a preliminary exploration, this study aimed to investigate the feasibility of M-mode ultrasound for the recognition of wrist and finger motions. Eight able-bodied subjects were tested for thirteen motions. The M-mode and B-mode ultrasound signals were acquired simultaneously from the forearm to ensure a fair comparison. Support vector machine (SVM) and back propagation (BP) neural network were adopted for classifying the wrist and hand motions. Regarding SVM classifier, the average classification accuracy (CA) of 13 motions across the 8 subjects was 98.83 % ± 1.03 % and 98.77 % ± 1.02 % for M-mode and B-mode, respectively. Regarding BP classifier, the average CA of M-mode and B-mode was 98.70 % ± 0.99 % and 98.76 % ± 0.91 %, respectively. There was no significant difference ( p > 0.05 ) on CAs between the M-mode and B-mode. Furthermore, M-mode showed potential superiority in feature space analysis. These results demonstrated the capability of M-mode ultrasound for the wrist and finger motion recognition. The outcomes showed the potential of M-mode ultrasound for its application in HMI.