Sensor-based gesture recognition with convolutional neural networks

Abstract

Sensor-based gesture recognition is an active research field of great significance with a wide range of applications in control systems for virtual reality, medical monitoring, and abnormal behavior determination. This problem has drawn the attention from both the academia and industry and many methods are proposed in the literature. Recently, deep learning has been widely applied in sensor-based gesture recognition and achieved good effects. In this paper, we proposed a classifier model based on Convolutional Neural Network (CNN) and applied it to EMG-based and smartphone-based datasets, respectively. For these two datasets, our model both achieved better classification accuracies than traditional machine learning models, with the results of approximately 97% and 72% accuracies, respectively. We also analyzed the effects of different parameters on the results of the proposed CNN model.