Research on sEMG-Based Gesture Recognition by Dual-View Deep Learning

In the field of human-machine interaction, gesture recognition using sparse multichannel surface electromyography (sEMG) remains a challenge. Based on Hilbert filling curve, dual-view multi-scale convolutional neural network (DVMSCNN) is designed to enhance performance in this paper. The consists two parts. first part, sEMG filled and obtained images time electrode domain are used as inputs block. second depth features learned by block fused classified “layer fusion” based view aggregation network. evaluation architecture four databases Ninapro-DB1, DB2, DB3 DB4 shows that DVMSCNN more than 7% accurate other state-of-the-art methods. When validated home-grown dataset, was able achieve rate 0.8848.