A multiple autocorrelation analysis method for motor imagery EEG feature extraction

Author

Xiangzhou Wang ; An Wang ; Shuhua Zheng ; Yingzi Lin ; Mingxin Yu

Author_Institution

Sch. of Autom., Beijing Inst. of Technol., Beijing, China

fYear

2014

fDate

May 31 2014-June 2 2014

Firstpage

3000

Lastpage

3004

Abstract

A novel multiple autocorrelation method for single trial EEG feature extraction was proposed. The time courses of ERD/ERS during motor imagery were investigated by calculating multiple autocorrelation before power spectrum analysis. Then the averaged power spectrums on specific frequency bands were sent to a K-nearest classifier to validate the separability between different classes. Compared with the result of power spectrum, the multiple autocorrelation performed better in attenuating noise and enhancing the separability between different classes with a small quantity of electrodes (C3 and C4). The maximum 90.0% accuracy tested on dataset of BCI-competition 2003 for motor imagery is achieved.

Keywords

correlation methods; electroencephalography; feature extraction; medical signal processing; signal classification; spectral analysis; BCI; ERD-ERS; K-nearest classifier; motor imagery EEG feature extraction; multiple autocorrelation analysis method; noise attenuation; power spectrum analysis; single trial EEG feature extraction; small electrode quantity; Accuracy; Correlation; Electrodes; Electroencephalography; Feature extraction; Noise; Spectral analysis; BCI; motor imagery; multiple autocorrelation; signal separability;

fLanguage

English

Publisher

ieee

Conference_Titel

Control and Decision Conference (2014 CCDC), The 26th Chinese

Conference_Location

Changsha

Print_ISBN

978-1-4799-3707-3

Type

conf

DOI

10.1109/CCDC.2014.6852688

Filename

6852688