Electroencephalogram identification method and device

Abstract

The invention provides an electroencephalogram identification method and a device. The electroencephalogram identification method comprises the following steps: firstly, forming a plurality of training sets by repeatedly extracting electroencephalogram samples from an original sample set; introducing a nuclear matrix for calculating a covariance matrix of electroencephalogram, thereby acquiring anaggregation space covariance matrix; decomposing characteristics and constructing a whitening matrix; utilizing the whitening matrix to convert the covariance matrix and performing characteristic decomposition, thereby acquiring a characteristic vector; constructing a space filter through the characteristic vector; extracting the characteristics of each electroencephalogram, thereby acquiring a classification model corresponding to a present training set; utilizing a plurality of classification models to confirm the category of to-be-classified electroencephalogram. According to the invention, the plurality of training sets is formed by repeatedly selecting the training samples, so that the individual difference of electroencephalograms can be effectively restrained and the characteristics with higher discriminability are extracted, so as to increase the recognition rate of electroencephalograms; and meanwhile, the nuclear matrix is introduced when the covariance matrix is calculated,so that the calculated amount can be reduced, the time consumption can be reduced and the treatment efficiency can be increased.

Description

Technical field [0001] The embodiments of the present invention relate to the technical field of brain electrical signal processing, and in particular, to a method and device for recognizing brain electrical signals. Background technique [0002] The BCI (Brain-Computer Interface) system is a system that uses electrophysiological signals to decode the user's intention into control commands to manipulate the device. According to the different ways of obtaining the user's thinking intention, EEG signals can be divided into Ecog, EEG, MEG, FMRI, etc. Among them, EEG signals have attracted wide attention because of their non-invasive and low-cost characteristics. The current BCI research based on EEG signals is mainly focused on the motor imagination EEG signals. Motor imagination is to generate related EEG signals by letting users "think". Research on motor imagination has shown that unilateral limb movement or imaginary movement can inhibit / enhance the rhythmic activity and power...

AI Technical Summary

Problems solved by technology

[0003] In the process of realizing the present invention, the inventors found that two major problems that cannot be avoided in EEG signal processing are: the accuracy of signal recognition and the time-consuming

However, the methods in the prior art do not solve these two problems well at the same time

Although CSP (common spatial pattern) has been proved to be an effective method to extract different types of motor imagery information in recent years, there are individual differences in EEG signals, and the CSP algorithm assumes that the collected EEG signals and brain sources There is a linear relationship between the signals, and the EEG signal collected from the top of the scalp is considered to be a nonlinear combination of brain-derived signals and noise sources, so linear CSP cannot fully explore the nonlinear structure of EEG EEG signals, and the recognition accuracy is not high

To improve the accuracy of EEG signal recognition, one way is to use large sample data for training and testing, such as using the DBN algorithm in deep learning and its derivative algorithms to build an EEG signal recognition model, which improves the EEG signal recognition rate. However, this method takes a long time, and the collection of large-sample EEG data also requires more sophisticated equipment and more manpower and material resources.

Images