Electromyographic gesture recognition method based on SeNet and gated time sequence convolutional network

Abstract

The invention relates to an electromyographic gesture recognition method based on SeNet and a gating time sequence convolutional network, and belongs to the field of electromyographic signal processing, and the method comprises the steps: S1, obtaining gesture electromyographic signal data, and dividing the data into a training set, a verification set and a test set; S2, preprocessing the gesture electromyographic signal data in the S1; S3, enhancing the gesture electromyographic signal data in the S2; S4, constructing a core layer; S5, constructing an attention mechanism layer; S6, constructing a complete model; and S7, inputting the data in the S3 into the complete model in the S6, training the model until a model loss function is not improved any more, and storing the model. According to the method, the data is enhanced and the myoelectricity data set is expanded, so that the recognition precision and generalization of the model are improved; and features between myoelectricity data channels are extracted by using SeNet, and the features are screened by using a gating time convolutional network, so that the gesture recognition precision of the network on myoelectricity signals can be effectively improved, and meanwhile, the requirements of real-time performance and high performance are met.

Description

technical field [0001] The invention relates to a myoelectric gesture recognition method based on SeNet and a gated time series convolution network, belonging to the field of myoelectric signal processing. Background technique [0002] Gesture recognition has become an important way of human-computer interaction, and the implementation of gesture recognition can be divided into vision-based gesture recognition and sensor-based gesture recognition. Among them, gesture recognition using electromyographic signals belongs to the latter. Compared with other gesture recognition methods, it has the advantages of wearability and low environmental interference. It has become one of the hot spots of human-computer interaction in recent years. Currently, gesture recognition based on electromyographic signals is mainly The method is divided into gesture recognition based on machine learning and gesture recognition based on deep learning. [0003] The process of gesture recognition base...

AI Technical Summary

Problems solved by technology

[0005] For the traditional machine learning method, the features for model training need to be manually extracted, and there will be problems of inaccurate feature extraction and complicated processing procedures. The deep learning method can effectively extract the features of the EMG signal by training the deep network, and has the advantages of The advantages of simple process, high robustness, and good recognition effect, but the deep learning method needs to set a suitable feature extraction layer, otherwise there will be problems such as low training accuracy, long training time, and gradient disappearance caused by too deep network

Images