2024-03-29T08:25:43Zhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_oaipmhoai:ipsj.ixsq.nii.ac.jp:000745802023-04-27T10:00:04Z01164:05352:06362:06450
定常視覚誘発電位に基づくBCI-振幅変化を利用した被注意刺激の判別に関する基礎的検討-Brain-computer interface based on steady-state visually evoked potentials — Fundamental study on classification of attended stimulus based on amplitude change —jpnNC研究会・一般講演(A)http://id.nii.ac.jp/1001/00074580/Technical Reporthttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=74580&item_no=1&attribute_id=1&file_no=1Copyright (c) 2011 by the Institute of Electronics, Information and Communication EngineersThis SIG report is only available to those in membership of the SIG.京都大学大学院工学研究科京都大学大学院工学研究科/日本学術振興会京都大学大学院工学研究科京都大学大学院工学研究科泉岡, 太輔笹山, 瑛由川ロ, 浩和小林, 哲生脳波 (electroencephalogram:EEG) に基づく BCI (brain-computer interface) では計測された EEG から特徴抽出を高精度に行うことが重要である.本研究では定常視覚誘発電位に基づく BCI に着目し,2 つの異なる周波数で点滅する視覚刺激のどちらに被験者が注意を向けているかを EEG から判別する新たな手法の提案を行った.先行研究では被験者は常に刺激に注意を向けているという前提で研究されており,疲労などによって注意がそれることを想定していないそこで短時間での特徴抽出を念頭に置き,計測された EEG に狭帯域バンドパスフィルタを施し,主成分分析と独立成分分析を用いて誘発成分とノイズを分離後,誘発成分の短い時間幅毎での振幅値にcommon spatial pattern を施して特徴を抽出し,サポートベクターマシンを用いて判別を行なった.全被験者平均の判別正答率は約 70% であり,BCI への実用に向けて提案した解析手法の有用性が示された.In BCI with EEGs, it is important to extract features precisely using signal processing techniques. In this study, we focused on a BCI based on SSVEPs (steady-state visually evoked potentials) and investigated the method to analyze SSVEPs and examined classification accuracy when subjects focused on one of two visual stimuli flickering at different frequencies. Most of previous studies performed on this kind of BCI assume that subjects continued to attend on stimuli and do not consider the effects on attention by fatigue. Here, we performed the feature extraction in the short time period. After applied band pass filter tuned at narrowband stimulus frequency for measured EEGs, we separated signals and noises by using principal component analysis and independent component analysis. Subsequently, we extracted features by applying common spatial pattern obtained to amplitude in each short analysis period of the ingredient corresponding to signals. Finally, the attended stimulus was determined by classifying the features with support vector machine. A classification accuracy rate in all subjects was about 70%. This demonstrates the feasibility of the proposal method as a BCI.AA12055912研究報告 バイオ情報学(BIO)2011-BIO-2516162011-06-162011-06-10