RESEARCH OF SACCADE-RELATED EEG: COMPARISON OF ENSEMBLE AVERAGING METHOD AND INDEPENDENT COMPONENT ANALYSIS Arao Funase 1,2 , Allan K. Barros 3 , Shigeru Okuma 2 , Tohru Yagi 4,5 , Andrzej Cichocki 1 1 Lab. for Advanced Brain Signal Processing, BSI, RIKEN, 2-1, Hirosawa, Wako, Saitama 351-0198, JAPAN {funase,cia}@bsp.brain.riken.go.jp 2 Lab. for Bioelectronics, Department of Information Electronics, Graduate School of Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Aichi 464-8603, JAPAN {arao@cmplx.cse,okuma@okuma.nuee}.nagoya-u.ac.jp 3 Dept. of Electrical Engineering, Fuderal University of Maranhao, UFMA, av. Dos Portugueses, Sao Luis, Ma, BRAZIL allan@biomedica.org 4 Nidek Vision Institute, Nidek 73-1, Hama-cho, Gamagori, Aichi 443-0038, JAPAN Thoru Yagi@nidek.co.jp 5 Lab. for Biologically Litegrative Sensors, BMC, RIKEN, 2271-130, Anagahora, Shimoshidami, Moriyama-ku, Nagoya, Aichi 463-0003, JAPAN yagi@bmc.riken.go.jp ABSTRACT Electroencephalogram (EEG) related to fast eye move- ment (saccade), has been the subject of application oriented research by our group toward developing a brain-computer interface(BCI). Our goal is to develop novel BCI based on eye movements system employing EEG signals on-line. Most of the analysis of the saccade-related EEG data has been performed using ensemble averaging approaches. However, ensemble averaging is not suitable for BCI. In order to process raw EEG data in real time, we per- formed saccade-realted EEG experiments and processed data by using the non-conventional Fast ICA with Reference sig- nal (FICAR). Using the FICAR algorithm, we was able to extract successfully a desired independent components(IC) which are correlated with a reference signal. Visually guided saccade tasks were performed and the EEG signal gener- ated in the saccade was recorded. The EEG processing was performed in three stages: PCA preprocessing and noise reduction, extraction of the desired IC using Wiener filter with reference signal, and post-processing using higher or- der statistics Fast ICA based on maximization of kurtosis. Form the experimental results and analysis we found that using FICAR it is possible to extract form raw EEG data the saccade-related ICs and to predict saccade in advance by 4[ms] before real movements of eyes occurs. For single trail EEG data we have successfully extracted the desire ICs with recognition rate 72%. 1. INTRODUCTION Brain-computer interfaces (BCIs) have been the subject of research efforts for several decades [1][2]. The capabilities of BCIs allow them to be used in situations unsuitable for the conventional interfaces. BCIs are used to connect a user and a computer via an electroencephalogram (EEG). The EEG is related to emotion, motion, and thought. There- fore, there is the potential that BCIs can be used to con- nect normal and mobility-impaired persons to computers in such a way that movement on the part of the user is not required. Moreover, the Quality of Life for severely handi- capped users is expected to be improved by using BCIs to connect these users to computers. EEG related to fast eye movement (saccade) have been studied by our group toward developing a BCI eye-tracking system that operates by using saccade-related EEG [3]. In previous research, EEG data was analyzed using the ensem- 867 4th International Symposium on Independent Component Analysis and Blind Signal Separation (ICA2003), April 2003, Nara, Japan