Emotion recognition method and device based on residual width network, medium and equipment

Abstract

The invention provides an emotion recognition method and device based on a residual width network, a medium and equipment. The method comprises the following steps: preprocessing electroencephalogram signal data; inputting into a residual width graph convolutional network; the residual width graph convolutional network comprises a graph convolutional network and a residual depth network; the graph convolutional network performs feature extraction on the irregular data and processes the irregular data into regular structured data; sending the structured data into each residual block of the residual depth network for feature extraction; the output of the graph convolution network and the output of each residual block of the residual depth network are connected in parallel to serve as output data of the residual width graph convolution network; and the width learning system maps the features to a width space, and final sentiment classification is realized through joint solution of feature nodes and enhanced nodes. According to the method, multi-layer features of all hierarchical structures are reserved, the accuracy and precision of classification tasks are improved, and the calculation cost is low.

Description

technical field [0001] The present invention relates to the technical field of EEG emotion recognition, and more particularly, to a method, device, medium and device for emotion recognition based on residual width network. Background technique [0002] Electroencephalography (EEG) is a commonly used physiological signal for emotion recognition, which has the characteristics of high accuracy and difficult to disguise. According to the frequency domain analysis of EEG, it is found that the five sub-bands of δ(1-4Hz), θ(4-8Hz), α(8-12Hz), β(13-30Hz), and γ(31-45Hz) are closely related to human’s frequency. Psychological activities are closely related. According to the link between frequency band characteristics and emotion, EEG signal analysis can be used as an effective way to identify human emotions. [0003] With the development of deep learning, the portability of EEG hardware, and the strengthening of dry electrode algorithms, the accuracy of EEG recognition of emotions ...

AI Technical Summary

Problems solved by technology

[0006] (1) Due to the shallow structure of the graph convolutional network, global features cannot be extracted; the update speed of the network is slow; in addition, the computational cost of the deep network is high, requiring longer training time and higher performance computing clusters;

[0007] (2) The performance of the width graph convolutional network at this stage has a significant decline after the depth is increased; the deep network will have network degradation problems and gradient dispersion / explosion problems when the number of layers is increased.

Images