Motor Imagery EEG Classification Method Based on Adaptive Decision Surface of LDA Classifier

Abstract

In order to solve the problem of reduced classification accuracy caused by the different spatial features of EEG motor imagery signals between the same subjects on different days, this paper proposes an EEG feature extraction and recognition method based on common spatial pattern (CSP) and decision surface adaptive linear discriminant analysis (DSALDA). The decision surface threshold of the linear discriminant analysis (LDA) classifier was updated by saving the CSP spatial feature of the different day's test data, and use parameters to control the proportion of CSP spatial features between the training data set and the different day’ test data set. The experiment collected EEG data of 24 subjects, each subject collected data on different days. The results show that the average accuracy of this method is improved by 6.35% compared with the CSP-LDA method, which effectively improves the accuracy and stability of the different day's motor imagery EEG classification. It has created conditions for the wide application of motor imagery brain-computer interface (BCI) system in the field of rehabilitation.