Enhancing accuracy of hand gesture recognition and its application by observing tendons using pneumatic mechanomyography (pMMG)

Abstract

This research addresses hand gesture recognition using pneumatic mechanomyography (pMMG). In order to enhance the accuracy of hand gesture recognition, observing the wrist tendon group was suggested as an alternative to previous forearm-worn observation. The observation of the tendon group, which is challenging to electromyography-based approaches, was solved by using pneumatic mechanomyography. We propose a novel fabrication method for pMMG sensors, which is simple, robust, and customizable. The method enabled observation on the forearm by reducing the minimum size of pMMG sensors. It is empirically shown that the observation has a linear relationship with finger flexion force. By using simple linear mapping, the suggested method achieved comparable or better accuracy in finger flexion force estimation compared to previous information-based optimization methods. In addition, the accuracy of hand motion recognition achieved 98.2% accuracy in the classification task, including 28 hand gestures using only pMMG sensors. Furthermore, 99.1% accuracy was achieved as a result of sensor fusion with EMG sensors. As a result, the suggested method achieved higher hand gesture classification accuracy compared to previous studies, including computer vision. In addition, it was experimentally shown that the proposed system operates stably even in dynamic applications in real environments.