Feature Extraction for Single Trial Record of Visual Mismatch Negativity by Use of Independent Component Analysis
Signal averaging method is usually utilized for extracting the characteristics of event related potentials (ERPs). However, the amplitude and duration of ERPs are not constant for each stimulus due to the fluctuation of the subjects state, accordingly the appropriate selection of available data is crucial for realizing the accurate averaging. Independent component analysis (ICA) is one of powerful tool for signal processing, and some application to analyze the neurological signal processing such as electroencephalogram (EEG), evoked potentials (EPs), ERPs and so on were done. In this study, ICA was applied to process the visual mismatch negativity (V-MMN) for extracting the features and for selecting the appropriate single trial data for averaging. From the grand average waveform of V-MMN, signal separation matrix was determined by use of ICA. Characteristic parameters for evaluating single trial data were calculated from the decomposed components of ERPs. Then, the available single trial data was selected based on the value of evaluation parameter. Waveforms of selective averaging method and conventional averaging method were compared and the effectiveness of the proposed method was examined.
Independent component analysis (ICA) visual mismatch negativity single trial data selective averaging.
T. Sugi K. Kimura S. Nishida T. Maekawa K. Ogata Y. Goto S. Tobimatsu M. Nakamura
Department of Electrical and Electronic Engineering, Saga University, Saga, Japan Department of Advanced Systems Control Engineering, Saga University, Saga, Japan Department of Information and Communication Engineering, Fukuoka Institute of Technology, Fukuoka, J Department of Clinical Neurophysiology, Kyushu University, Fukuoka, Japan Department of Occupational Therapy, International University of Health and Welfare, Fukuoka, Japan
国际会议
北京
英文
1483-1487
2007-05-23（万方平台首次上网日期，不代表论文的发表时间）