Motor imagery electroencephalogram signal classification method based on channel attention and multi-scale time domain convolution

Abstract

The invention discloses a motor imagery electroencephalogram signal classification method based on channel attention and multi-scale time domain convolution, and belongs to the field of computer software. In order to solve the problem that feature extraction is difficult due to the fact that the signal-to-noise ratio of electroencephalogram signals is low, the invention provides an improved network model based on EEGNet, namely 'MCA-EEGNet' for short. Firstly, a parallel multi-scale time convolution layer is used for replacing a common convolution layer in an EEGNet model, so that feature extraction is better carried out, and the classification accuracy is improved. And meanwhile, a channel attention module ECA is added, so that channel information with high relevancy with input data is more concerned during network training, and the robustness of the model is further improved. Compared with an EEGNet model, the classification method provided by the invention can more effectively improve the feature extraction and classification performance of the motor imagery electroencephalogram signals.

Description

technical field [0001] The invention discloses a motor imagery EEG signal classification method based on channel attention and multi-scale time-domain convolution, which can be used to identify motor imagery body parts and belongs to the field of computer software. Background technique [0002] In recent years, EEG-based brain-computer interface (Brain–computer interfaces, BCI) technology has developed rapidly. Brain-computer interface is a real-time communication system connecting the brain and external devices. It has important research significance and great application potential in the fields of biomedicine, neurorehabilitation and artificial intelligence. It realizes a new way of direct communication between the brain and the outside world. Human-computer interaction mode. There are three main ways for brain-computer interface technology to collect data from the brain: invasive, semi-invasive, and non-invasive. EEG-based BCI is non-invasive and is the most commonly us...

AI Technical Summary

Problems solved by technology

Aiming at the problem of difficult feature extraction due to the low signal-to-noise ratio of EEG signals, a parallel multi-scale temporal convolutional layer is used instead of the ordinary convolutional layer in the EEGNet model to better perform feature extraction and improve classification accuracy; at the same time, add The channel attention module ECA makes the network training pay more attention to the channel information with high correlation with the input data, further improving the robustness of the model

Images