Language Recognition System Based On Bottleneck Features

Spoken language identification(LID)is all automatic process that aims to determine the language identity of a given speech segment.With the rapid progress of the communication techniques,LID plays a more and more important role for multilingual speech processing system.A significant progress in LID technique has been witnessed in the past few decades.Currently,the front-end features for LID system include spectral and phonetic ones.This is insufficient for challenging short duration and confusable dialect recognition.This thesis investigates the use of deep learning theory for extracting robustness feature and improving model capability for language recognition.This article mainly around the Deep learning in the language recognition application,based on Deep Neural Network(Deep Neural Network,within DNN)to extract the phonemes associated Deep Bottleneck Features(Deep Bottleneck Feature,DBF)of the language recognition method to do the related research.The DBF can effectively inhibit the acoustic features of the underlying language is independent of the noise,especially the channel difference,the difference of the speaker,the noise factor of the difference of background noise,and can be the underlying acoustic features and the physical significance of the underlying acoustic unit.Experimental results show that using DBF features combining the difference modeling(the Total Variability,TV)language recognition method of DBF-TV,can effectively promote the performance of the language recognition,especially greatly improved the easy mixed languages and dialects,short speech language recognition performance.