3D-mechanomyography: Accessing deeper muscle information non-invasively for human-machine interfacing

Various findings on the study of mechanomyography (MMG) have emphasized the importance of contact force between the sensor and the skin in the evaluation of MMG activity. Although these studies suggest that contact pressure may have some overall alteration on the quality of MMG signal, they do not provide insight about how to use different pressure levels to extract more useful information regarding human motor intent. In this paper, we detail the design of a custom-made MMG sensor, which allows for systematic evaluation of different levels of contact force between the sensor and the skin. We use 3 increasing levels of force to evaluate the classification of 6 different hand gestures (Flexion, Extension, Pronation, Supination, Adduction, Abduction). Four MMG channels were positioned around the forearm and placed over the flexor carpi radialis muscle, brachioradialis muscle, extensor digitorum muscle, and flexor carpi ulnaris muscle. A total of 252 spectrotemporal features were extracted and used to train a linear support vector machine (SVM). The average classification accuracies for the 3 increasing levels of contact force for all the participants are 57.68%, 57.31%, 65.27%, respectively. Our preliminary results show that contact pressure has a clear effect on the discriminability power of MMG signals. As shown in the results, the classification performance tends to increase as the contact pressure increases; however, further experiments are required to quantify the extent of this effect during static and dynamic actions.