Analysis of Blind Source Separation Techniques for Eye Artifact Removal

• Published in: Wireless Networks and Computational Intelligence pp. 340 - 349
• Main Authors: Aspiras, Theus H., Asari, Vijayan K.
• Format: Book Chapter
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg 2012
• Series: Communications in Computer and Information Science
• Subjects: Electroencephalography; Electrooculography; Eye Artifact Removal; Independent Component Analysis; Strength of Eye Blink
• ISBN: 3642316859; 9783642316852
• ISSN: 1865-0929; 1865-0937
• DOI: 10.1007/978-3-642-31686-9_40

Evaluation of several different eye artifact removal techniques for electroencephalographic data is presented in this paper. Data is taken from an emotion recognition experiment, in which subjects undergo five different emotions (joy, sadness, disgust, fear, and neutral). Preprocessing for the EEG Data includes filtering with a Butterworth band-pass filter and a 60Hz notch filter. Three different types of eye artifact removal techniques are explored using the preprocessed data: EOG based linear regression, Principal Component Analysis, and Independent Component Analysis. All techniques used electrooculographic (EOG) data to determine the criteria for feature extraction and removal. Evaluations from our experiments show that all techniques significantly reduce the effects of eye blinks and eye movements in the EEG. The developed metric used in experimentation shows that Independent Component Analysis reduced eye artifacts the best while keeping EEG portions unchanged (Average SSE of 0.1126 for clean EEG portions).