Affiliation of Author(s):计算机科学与技术学院/人工智能学院/软件学院

Journal:INTELLIGENT DATA ENGINEERING AND AUTOMATED LEARNING - IDEAL 2017

Key Words:Brain computer interface Motor imagery Common spatial patterns Artifact removal

Abstract:EEG reflects the strength of the neuronal activity in the brain. Since EEG signal is weak, noisy and mixed with a large number of artifacts, which causes interference to the processing and identification of the EEG signal. Using EEG related pretreatment can effectively remove artifact, noise, and improve EEG signal-noise ratio and efficient, which provides more accurate data for feature extraction and classification. In this paper, we introduce several methods including PCA, ICA and CSP. Based on these methods, the complete process of EEG signal de-noising, feature extraction and classification are established, which can complete the classification and recognition of the motor imagery signals. We use a combination of a lot of pretreatment methods to analysis and process motor imagery of EEG and propose an improved algorithm named CS-CSP. The experimental results show that the Chebyshev type II filter is superior to the conventional pre-treatment methods and the recognition accuracy of CS-CSP is higher than CSP.

ISSN No.:0302-9743

Translation or Not:no

Date of Publication:2017-01-01

Co-author:朱延龙,王钟毓,代成龙

Correspondence Author:Pi Dechang