Gesture recognition based on improved convolutional neural network based on self-attention mechanism

This article presents the original 10-channel surface electromyography (sEMG) signals and analyzes their specific roles in six muscle groups. It introduces the CNN architecture and provides details of experimental parameters. The model training is conducted, with accuracy (Acc) and loss as the evaluation metrics. Three different approaches are attempted to improve the model’s performance, and the results are discussed. The experimental findings demonstrate that CNNbased models have the potential for further improvement, but the sole introduction of attention mechanisms does not lead to a significant enhancement in model performance.In summary, this paper provides insights into the analysis of sEMG signals and the application of a CNN-based model for muscle group recognition. It highlights the challenges in improving model performance and emphasizes the need for more advanced techniques beyond attention mechanisms. This research contributes to the field of electromyography and deep learning for muscle activity recognition.