EMGおよびMMG信号を用いた義手制御のためのハンドパターン認識法Signals

抄録

Myoelectric control system is a fundamental component of modern prosthetic devices, which uses the myoelectric signals from an individual’s muscles to control the prosthetic’s movements. Currently, surface electromyography (EMG) is the most popular myoelectric control used in prosthetics to detect electrical signals generated by muscles. (MMG) is a less known technique that involves detecting and interpreting mechanical vibrations generated by muscle contractions. This technique requires no electrical connection to muscle tissues, thus making them more resistant to noise interference and is also a cheaper alternative. This study intends to investigate an alternative method to classify hand gestures for prosthetic control by simultaneous acquisition of EMG and MMG signals. The sensors used are two Myoware Muscle sensors to record EMG signals and two MPU6050 accelerometers to record MMG signals. The hand gestures measured are gripping, extension, flexion, supination, and pronation. Pattern recognition using long short-term memory (LSTM) method is used to classify hand gestures.

With the resulting overall accuracy of 86.8%, this study confirms that the simultaneous acquisition of EMG and MMG signals has the potential for anthropomorphic prosthetic control.