Multi-stream CNN-SVM Hybrid Model for Gesture Recognition based on sEMG Signals

Gesture recognition has various applications in fields such as prosthetic control and human-machine interaction. In recent years, various gesture classification models based on surface electromyography (sEMG) signals have been proposed. Feature extraction of sEMG signals is discussed in this paper, and four models are used for recognizing gestures based on sEMG signals, which are convolutional neural networks (CNN), Multi-stream CNN, support vector machines (SVM) and Multi-stream CNN-SVM. The classification results show that Multi-stream CNN and SVM have better recognition performance than classical CNN. In addition, a Multi-stream CNN-SVM hybrid model is proposed. Compared with relevant research by other scholars, our proposed hybrid model performs better in gesture classification problems, with the highest classification accuracy of 97.7%.