CAPG-MYO: A Muscle-Computer Interface Supporting User-defined Gesture Recognition

Abstract

The recent progress in sEMG-based hand gesture detection has developed a set of predefined hand gestures for interaction. However, customized hand gestures are less concerned due to the lack of supporting tools for training alternative hand gestures. To fill the gap, we present a training system, called CAPG-MYO, for user-defined hand gesture interaction. An armband named CAPG was used to simultaneously capture sEMG and IMU signals from the participants to construct a small-scale dataset with the customized hand gestures. To predict the customized hand gestures, we developed a multiview convolutional neural network to handle the dataset and consequently shaped a gesture recognition model that effectively transferred a model of 8-hand gesture recognition into a model of the customized hand gesture recognition. We conducted technical validation of the system and the results demonstrate the system's accuracy of hand gesture recognition, which reached 84.71% after a 2min training.