Surface electromyography (sEMG) signals are the sum of action potentials emitted by many motor units; they contain the information of muscle contraction patterns and intensity, so they can be used as a simple and reliable source for grasping mode recognition. This paper introduces the InRes-ACNet (inception–attention–ACmix-ResNet50) model, a novel deep-learning approach based on ResNet50, incorporating multi-scale modules and self-attention mechanisms. The proposed model aims to improve gesture recognition performance by enhancing its ability to extract channel feature information within sparse sEMG signals. The InRes-ACNet model is evaluated on the NinaPro DB1 and NinaPro DB5 datasets; the recognition accuracy for these datasets can reach 87.94% and 87.04%, respectively, and recognition accuracy can reach 88.37% in the grasping mode prediction of an electromyography manipulator. The results show that the fusion of multi-scale modules and self-attention mechanisms endows a strong ability for the task of gesture recognition based on sparse sEMG signals.

Surface electromyography (sEMG) signals reflect the local electrical activity of muscle fibers and the synergistic action of the overall muscle group, making them useful for gesture control of myoelectric manipulators. In recent years, deep learning methods have increasingly been applied to sEMG gesture recognition due to their powerful automatic feature extraction capabilities. sEMG signals contain rich local details and global patterns, but single-scale convolutional networks are limited in their ability to capture both comprehensively, which restricts model performance. This paper proposes a deep learning model based on multi-scale feature fusion—MS-CLSTM (MS Block-ResCBAM-Bi-LSTM). The MS Block extracts local details, global patterns, and inter-channel correlations in sEMG signals using convolutional kernels of different scales. The ResCBAM, which integrates CBAM and Simple-ResNet, enhances attention to key gesture information while alleviating overfitting issues common in small-sample datasets. Experimental results demonstrate that the MS-CLSTM model achieves recognition accuracies of 86.66% and 83.27% on the Ninapro DB2 and DB4 datasets, respectively, and the accuracy can reach 89% in real-time myoelectric manipulator gesture prediction experiments. The proposed model exhibits superior performance in sEMG gesture recognition tasks, offering an effective solution for applications in prosthetic hand control, robotic control, and other human–computer interaction fields.