Research On The Chinese Speech Pronunciation System Of DIVA Model

DIVA model is a kind of mathematical model which describes the role played in human brain refer to speech production and speech understanding area. The model can simulate the movement of how human brain to control speech organ, so as to simulate the pronunciation of phoneme, word and sentence of human. So far, DIVA model is the neural network model which only uses pseudo-inverse control scheme and it also has biological significance. Therefore, DIVA model has guiding significance to work out speech brain-machine interface system which is able to substitute human pronunciation through acquiring signal giving by human brain. The main work of this paper:1.Makes study on the predication and definition of DIVA model as well as designs fMRI experiment associated with Chinese speech. 2 Modifies brain region of interest of DIVA model against the pronunciation process of Chinese speech. 3.Extracting the electroencephalogram signal produced in these regions. Due to electroencephalogram signals obtained through noninvasive electroencephalogram signal acquisition scheme have more interference of electronystagmography and electromyogram, etc., so we use independent component analysis to exclude noise signal of electroencephalogram signal, then adopt wavelet packet decomposition and common spatial pattern to conduct feature extraction for electroencephalogram signal, use support vector machine classifier to classify and work out a brain-machine interface system aiming at Chinese speech pronunciation.Experimental result shows that using Chinese speech brain-machine interface system designed in this paper can reach a good performance, classification accuracy of support vector machine model performed on Chinese speech electroencephalogram signal achieves 75%, which reaches the classification accuracy of English speech brain-machine interface designed by professor Gunther using intrusive mode electroencephalogram signal collection method, so as to put in evidence that this study provides a feasible program for the design of Chinese speech nervous analytical system. In addition, the conclusion drawn from this experiment also verifies the activity predication of DIVA model on brain region in the pronunciation process. Therefore, this paper lays a good foundation on the Brain-Computer Interface and production of Chinese speech in the future.