Based On The Research And Design Of Embedded Appliances Words Speech Recognition System

Speech Recognition (SR) is such a kind of technology that deals with speech signal with computer and recognizes them to symbol sequence. The technology of small or medium vocabulary speech recognition, such as voice command recognition, keyword spotting and continuous digit string speech recognition, are of great importance in the application-oriented study of speech recognition.The paper designs a embedded speaker-independent keyword spotting system which is used to spot and recognise keyword in human's control speech to their household appliances. It will serve the executive system. We can sure that it will greatly improve automatization of family life. The paper works mainly at two aspects: one is to research every component of speech keyword spotting system, put forward and realize a keyword spotting project which aims at embedded application; the other is to research and realize keyword spotting based on DSP platform.The paper simplifies the continuous HMM in order to suit embedded application. Alternate model strategy is proposed to realize keyword spotting. The experiment shows that it's effective. System's correct detection rate reaches 83.51% when the number of alternate model is 5, and the detection performance is good. An continuous speech endpoint detection method based on twice syncopation is proposed, and it works well at detecting syllable in chinese continous speech. A speech database for training, testing and adaptation is created. Speaker adaptation which uses given specker's speech sample to modulate the Speaker-Independent model has been researched, in order to improve system's recognition performance to the given speaker. The paper researches interface technology of DSP and realizes that DSP collects speech from a micphone and do the keyword spotting.