基于颅内脑电的汉语语音基本单元神经编码与解码研究

项目名称： 基于颅内脑电的汉语语音基本单元神经编码与解码研究

项目编号： No.61473169

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 其他

项目作者： 洪波

作者单位： 清华大学

项目金额： 81万元

中文摘要： 大脑处理语音信息的基本单元尚不清楚，但认识这一机制对于理解语言的神经机理以及推动计算机语音识别技术、脑机接口技术的发展都具有重要意义。由于语音信息具有快速时变特性，加之言语处理脑区的空间特异性强，本课题采用神经外科颅内脑电技术获取大脑在汉语语音刺激下的精细时间和空间响应。为建立语音信息神经表征模型，本课题集中在语音信息处理的中枢和关键层次，探索音素层次以及更高一级的音节层次的神经响应时空频特性，确定其基本处理单元，以及神经编码空间的维度和结构。特别地，本课题针对汉语语音体系的特点，研究声母、韵母和声调不同的群体编码机制，并提出汉语音节层次的重要性。在编码研究的基础上，我们综合不同层次的神经表征模型，通过贝叶斯解码构建基于颅内脑电的汉语语音脑机接口系统。

中文关键词： 颅内脑电；语音感知；神经解码；脑机接口

英文摘要： What is the basic unit of speech processing in the brain is still poorly understood. However, answering this question is of great importance for revealing the neural mechanism of speech information processing, and for promoting the development of automatic speech recognition and brain-computer interfaces technologies. Considering the fast-changing nature of speech acoustic information and the distributed functional network of speech processing, intracranial EEG technology is used to obtain a fine spatio-temporal resolution of the neural responses to Chinese speech. To establish the neural representation model, we focus on the key stages of speech information processing at both the phoneme and syllable levels, and explore the temporal, spatio-spectral patterns of the elicited neural responses. These data are then analyzed to determine the basic speech processing unit and the corresponding neural coding scheme. Specifically, vowel, cosonant and tone are included when addressing the phoneme level to investigate the speciality of Chinese language. Based on these findings, a brain-computer interface for Chinese speech communication is constructed based on a Baysian decoding approach synthezing the neural representation models of both the phoneme and syllable levels.

英文关键词： Intracranial EEG;Speech Perception;Neural Decoding;Brain Computer Interface