Scalp EEG feature extraction and classification method based on end-to-end convolutional neural network

Abstract

The invention discloses a method for extracting and classifying scalp EEG features based on an end-to-end convolutional neural network. Data enhancement is performed on training data, and then the enhanced training data is used to train the convolutional neural network; the data to be detected is input into the convolutional neural network for further processing. The feature extraction and classification steps are: S1. Filter and process the original scalp EEG signal with a band-pass filter to obtain the signal x θ 、x μ and x β ;S2, pair signal x θ 、x μ and x β Perform multi-scale temporal convolution and spatial convolution to extract features; S3, perform pooling operation on the feature map output by the convolutional layer; S4, perform feature fusion after pooling, and then send it to the fully connected layer to perform abstract feature extraction on the input Integration; S5, the output of the fully connected layer is sent to the softmax layer for classification. The present invention applies a brand-new data enhancement technology in the training stage. In the test stage, the data is first passed through the filter bank, and then input into multiple convolutional neural network branches to perform multi-scale convolution operations, which reduces the over-fitting phenomenon and improves the classification. Accuracy.

Description

technical field [0001] The invention relates to the technical field of brain-computer interface, in particular to a scalp EEG feature extraction and classification method based on an end-to-end convolutional neural network. Background technique [0002] A brain-computer interface is a communication system that helps people use their brains to control and use external devices without the involvement of peripheral nerves and muscles. This brings hope to those patients with normal thinking but severe neuromuscular damage. The brain-computer interface can regain the ability to move or communicate with the environment and improve the quality of life of patients. Although the original purpose of the brain-computer interface system was to provide a new way of communication for patients with complete paralysis, it has shown great potential in the field of entertainment games in recent years. There have been many studies on the application of brain-computer interfaces to some popular...

AI Technical Summary

Problems solved by technology

[0004] However, the currently proposed method based on deep neural network still has the following problems. First, due to the great success of convolutional neural network (CNN) in computer vision, many studies have applied CNN to the field of EEG motion image recognition, but in The CNN used in existing work has a fixed convolution scale (that is, a fixed convolution kernel size), but for different people, the convolution scale with the highest classification accuracy is not the same, so the fixed convolution scale limits the classification accuracy. Second, due to the complexity of the deep neural network and the huge amount of parameters, network training requires a large amount of data, and high-quality EEG data needs to be carried out in an environment with low electromagnetic interference, which is a very time-consuming and labor-intensive task. Limited amount, easy to overfit

Images