DIVA (Direction Into Velocities of Articulators)模型是一种描述人脑中涉及语音生成和语音理解区域所发挥的作用的数学模型, 能对发音过 程进行模拟, 对语音脑机接口系统的设计具有指导意义. 文章根据DIVA模型的定义和相关研究结论, 对人在发音过程中的脑电信号进行了处理, 首先利用小波包将脑电信号进行特征提取, 之后使用SVM (Support Vector Machine)分类器进行分类. 结果表明, 该方法对发音过程的脑电信号特征提取和分类效果较好, 识别率达到70\%, 为基于DIVA模型的语音脑机接口系统设计提供了一 种思路, 此外, 实验的结论也印证了DIVA模型对于发音过程大脑区域激活情况的预测.

DIVA (Direction Into Velocities of Articulators) model shows the role of the brain involved in speech production and understanding by the region, it can simulate the pronunciation process and play a key role in speech BCI design. In order to realize the speech BCI system, we used wavelet packet decomposition and SVM (support vector machine) in EEG signal processing during the speech in this study according to the forecast of DIVA model and the conclusions of DIVA-related researches. Result shows that the eigenvector extracted by the method worked out fine in the classification of EEG signal and the accuracy rate of 70\%, it provided a new way to design a speech-related BCI system. In addition, the conclusions of this study also proved the forecast of brain activation during the speech of DIVA model.