A novel approach of FMG sensors distribution leading to subject independent approach for effective and efficient detection of forearm dynamic movements

Abstract

In force myography (FMG) limb movement is detected by measuring muscle contraction intensity in terms of forces. In this work, FMG is used to detect forearm movements i.e. flexion, extension, pronation, supination and steady-state. In FMG, sensors are conventionally located at supinator/pronator muscles to detect forearm pronation/supination movement. In this paper, a new approach to detect forearm movements i.e. pronation/supination also including flexion/extension and steady-state posture, is proposed. This is achieved by designing a sensor distribution pattern on upper arm muscles and supported by bagged tree ensemble classification algorithm a unified classification model is obtained and tested on multiple subjects. Performance of the method is evaluated using accuracy, precision, and recall. Results have shown that with the proposed method a unified model can be developed for detecting forearm movements. An average of 0.91, 0.93 and 97.3% precision, recall and accuracy, respectively, is achieved.

Keywords

• Force myography
• Forearm movement detection
• Cross-subject model
• Machine learning