CSP (Common Spatial Patterns) and cross-correlation based motor imagery electroencephalogram classification method

Abstract

The invention discloses a CSP (Common Spatial Patterns) and cross-correlation based motor imagery electroencephalogram classification method which mainly aims at solving the problems that the adaptability to individuals is poor and the classification accuracy difference is large in the prior art. The CSP and cross-correlation based motor imagery electroencephalogram classification method comprises the implementation steps of step 1, collecting electroencephalogram signals and obtaining a training set and a test set; step 2, extracting a common spatial feature of the training set in CSP, simultaneously extracting a cross-correlation feature of the training set through a cross-correlation function and combining into a training feature vector of the training set through the two features; step 3, extracting a test feature vector of the test set by the same method in the step 2; step 4, performing training on a support vector machine through the training feature vector, obtaining a support vector machine classifier and performing classification of imagery movement tasks on the test feature vector through the support vector machine classifier. The effect of the individual difference on the classification result is reduced, the classification accuracy is improved, and the CSP and cross-correlation based motor imagery electroencephalogram classification method can be applied to the control on electroencephalogram products comprising motor imagery BCI (Brain Computer Interface) on-line systems.

Description

technical field [0001] The invention belongs to the field of information technology, and further relates to a method for classifying four types of motor imagery EEG signals of left hand, right hand, foot and tongue, which can be used for the control of EEG products with motor imagery brain-computer interface BCI online systems such as automatic wheelchairs. Background technique [0002] According to neuroanatomy, the surface layer of the human brain is called the cerebral cortex. The cerebral cortex can be divided into several functional areas, and different areas are in charge of and regulate different functions and mechanisms of the body. According to the spatial location, the cerebral cortex is divided into several lobes, including the frontal lobe, parietal lobe, occipital lobe, and temporal lobe. Among them, the frontal lobe mainly determines personality, controls emotions, and distinguishes right from wrong; the parietal lobe can feel touch and control body movements,...

AI Technical Summary

Problems solved by technology

However, as documented in "LI Ming-Ai Lin Lin Yang Jin-Fu, 'Adaptive Feature Extraction of Four-Class Motor Imagery EEG Based on Best Basis of Wavelet Packet and CSP,' Electric Information and Control Engineering (ICEICE), International Conference, 3918-3921 (2011)", however, this common space mode CSP algorithm has the following disadvantages: 1. It is sensitive to noise, that is, when the noise is relatively large, the quality of the extracted feature vector will be greatly reduced. 2. Different individuals cannot adapt to their individual differences. Since each individual performs the same imaginary movement, the EEG signals are different and have differences, and the common space mode CSP is not very good. To adapt to this difference, it shows that the classification accuracy rate of some individuals is low

Images