Reaearch On Several Problems Of Spech Production And Acquisition Based On DIVA Model

In order to study human cognitive development, we need a system that has the samephysiological capacities and the same ecological setting as an infant,DIVA(Directions Into ofArticulators) model is this kind of system.The DIVA model is a kind of mathematical model thatabout the processing procedure after speech production and acquistion.It is mainly used to simulateand discribe some related functions involve speech production and speech understanding areas inbrain.It is a kind of adaptive neural network that in order to produce words,syllable or phoneme tocontrol simulation vocal tract movement.Since 1995 to now,it appears different versions of DIVA model,but they all exist someimperfections more or less,two of them are as followed:First is the question of effect of speakingstyle and speaking rate on speech production.Secondly is the question of phoneme initialization inDIVA model.This paper discuss and study these two questions above,in section three we propose a formantcontour model based on DIVA model to study the effect of speaking style and speaking rate onspeech production.The experimental result indicate that the slope depends on speaking style for thevowels at onset, and depends on speaking style and the speaking rate for some vowels at offset.Insection four we propose a method of automaticly acquire speech representation to resolve thesecond question. The experimental result indicate that apply this method in DIVA model couldmake the model be more natural,and make the study of some initial process of speech acquisitionprocess be better. Finally, we summarize the research contents, and point out the subsequent studydirection, and extend thinking for futher study.