Quick Search Adv. Search

Journal of Bionic Engineering ›› 2022, Vol. 19 ›› Issue (3): 700-708.doi: 10.1007/s42235-022-00171-7

Previous Articles     Next Articles

Comparing EMG Pattern Recognition with and Without Hand and Wrist Movements

Lizhi Pan1, Kai Liu1, Kun Zhu1, Jianmin Li1   

  1. 1 Key Laboratory of Mechanism Theory and Equipment Design of Ministry of Education, School of Mechanical Engineering, Tianjin University, Tianjin 300072, China
  • Received:2021-09-22 Revised:2022-01-04 Accepted:2021-01-21 Online:2022-05-10 Published:2022-05-04
  • Contact: Jianmin Li E-mail:mjli@tju.edu.cn
  • About author:Lizhi Pan1, Kai Liu1, Kun Zhu1, Jianmin Li1

Abstract: Electromyography (EMG) pattern recognition has been widely employed for prosthesis control. Several studies demonstrated that amputees had poorer performances of EMG pattern recognition when compared to able-bodied individuals. Several factors, such as the muscle weakness and atrophy of residual limbs, the length of residual limbs, and the decrease of the affected side's motor cortex, had been studied to improve the performance of amputees. However, there was no study on the factor that the absence of joint movements for amputees. This study aimed to investigate whether the hand and wrist joint movements had effects on the EMG pattern recognition. Ten able-bodied subjects were tested for 11 hand and wrist gestures with two different gesture modalities: hand and wrist joints unconstrained (HAWJU) and constrained (HAWJC). Time-domain (TD) features and Linear Discriminant Analysis (LDA) were employed to compare the classification performance of the two modalities. Compared to HAWJU, HAWJC significantly reduced the average Classification Accuracy (CA) across all subjects from 95.53 to 85.52%. The experimental results demonstrated that the hand and wrist joint movements had significant effects on EMG pattern recognition. The outcomes provided a new perspective to study the factors affecting EMG pattern recognition.

Key words: Electromyography (EMG), Pattern recognition, Classifcation accuracy (CA), Joint movements