Motor Imagery Analysis Based on Filter Bank Common Spatial Pattern

This paper analyzes the EEG signals of left and right hand motor imagery and uses Filter Bank Common Spatial Pattern to extract features from EEG signals. Compared with Discrete Wavelet Transform, Autoregressive mode, Power Spectral Density and Common Spatial Pattern, it is found that FBCSP can significantly improve the recognition rate. While extracting the features of sub-frequency bands, this paper also considers ERD and ERS in motor imagery and the distribution of signal energy in different frequency bands. It is found that during the time of the motor imagery, the energy changes are generally concentrated on the μ rhythm, and the motor sensory area of the cerebral cortex has the nature of contralateral reflection of left and right hand motor imagery. Researchers generally use Support Vector Machine classifiers for pattern recognition classification, but the classification effect of a traditional SVM classifier is not ideal. Therefore, the paper uses the integration algorithm AdaBoost and algorithm Gradient Boosting for classification. Compared with Adaboost, the error rate of Gradient Boosting algorithm drops more sharply with the increased number of iterations, and the accuracy of Gradient Boosting algorithm is higher.