基于汉语音位发音想象的脑机接口研究

发布时间：2018-05-30 16:17

本文选题：脑机接口 + 运动想象　；参考：《中文信息学报》2014年05期

【摘要】：该文提出了一个基于汉语音位发音想象的脑机接口系统框架,使得受试者使用脑机接口系统时能更加自然和流畅。三名受试者参与了本实验研究,实验过程中受试者被要求想象四个汉语元音和四个辅音音位的发音部位及语音发音,以及一个不作想象任务的控制条件,同时记录其脑电数据。在数据处理阶段,本文对采集到的头皮脑电数据进行了频域、时域、空域分析,以提取出音位发音想象效应最优化的特征向量用于提高每两个条件间的配对分类效果。实验结果表明,音位发音想象效应的最优脑电频段为2~10Hz,时段为刺激呈现后300~500ms,头皮空间分布主要集中在感觉运动皮层区域。音位发音想象任务和控制条件相比具有较高的分类正确率,最高可达83%,为基于音位发音想象的汉语脑机接口系统研究提供了理论基础。此外,刺激材料间的Jaccard距离和分类正确率的高度相关性表明,音位发音想象任务可被视为复杂的发音器官运动想象任务,并且可由人脑感觉运动皮层区域的脑电信号来解码预测。
[Abstract]:In this paper, a framework of brain-computer interface system based on Chinese phonetic phonetics imagination is proposed, which makes the subjects more natural and smooth when using the brain-computer interface system. In this study, three subjects were asked to imagine the pronunciation position and pronunciation of four Chinese vowels and four consonants, as well as a control condition without imagination, and to record the EEG data at the same time. In the stage of data processing, the scalp EEG data were analyzed in frequency domain, time domain and spatial domain to extract the feature vector of phonetic imagination effect optimization to improve the matching classification effect between each two conditions. The experimental results showed that the optimal EEG frequency band of phonetic imagination effect was 210Hz, the time interval was 300 ~ 500ms after stimulation, and the spatial distribution of scalp was mainly concentrated in the sensorimotor cortex. The task of phonetic imagination has a higher classification accuracy than that of the control condition, which can reach 83%, which provides a theoretical basis for the research of the Chinese brain-computer interface system based on phonetic imagination. In addition, the high correlation between the Jaccard distance between stimulative materials and classification accuracy indicates that phonetic imagination can be considered as a complex task of phonetic organ motion imagination. Furthermore, the prediction can be decoded by the EEG signal in the sensorimotor cortex of the human brain.
【作者单位】：清华大学人文学院计算语言学实验室;清华大学人文学院心理学与认知科学研究中心;
【基金】：国家自然科学基金(61171114) 清华大学自主科研项目(20111081023&20111081010)
【分类号】：TP334.7;TN912.3

【相似文献】