Research And Development Of Android-Based Uyghur Speech Assistant Application Software

With the rapid development of modern science and technology,Speechrecognition had made many achievements.Especially, in recent years, with the wideapplication of intelligent mobile phones, personal handheld computers(Pad) and otherintelligent mobile devices, speech keyword spotting technology embodies its moreimportant role. Speech as the most direct and most convenient way of interaction,human-computer interaction is the mankind for a long time, but also the final goal ofspeech recognition technology. Therefore, based on the study of Uyghur languageKeyword Recognition and based to the “Aitalk” of iFLYTEK Company, realized theUyghur speech assistant application software on theAndroid.The main research contents of this paper are as follows:Firstly,In this paper,Acoustic model we using the Uyghur reading corpus totraining and taking into account the context, the single phoneme model originallyannotation using problem sets into sharing three triphone model, and the state-tied bythe decision tree.For the establishment of language model, we using the approach ofrules based, namely the establishment of slot grammar network. Slot grammar files,we construct a total of15Slot and three rules, namely three <action>, then usingHparse of HTK to derive the syntax network, build language models for subsequentcommand word recognition. Recognition will eventually use the slot grammar rules toguide the command word.Then, Use the Acoustic model and Language model testing for test datarecognition by HTK’s HVite.The end,On the basis of acoustic and language models have been training andbuilding, reference iFLYTEK of “Aitalk4”, designed Uyghur Command WordRecognition System.Uyghur Command Word Recognition System has achieved the eight main functions. For respectively calling, message sending, open the application,music search, Webpage navigation, get the test information and the latest news,weather query functions. The system testing actual.