Based On Telephone Voice Recognition System Of Uyghur Keywords

This paper gives a small vocabulary, keyword recognition identification system based on HTK. It is the most popular speech recognition tools in the word now. Keywords recognition is an important area of speech recognition research,Its purpose is to detect and confirm the given a number of specific keywords from the continuous speech. The recognition of key-words is very important in the telephone monitoring and voice retrieval.In this paper, we do the following task with their pronunciation and grammar Uyghur characteristics:1. We established a Uyghur continuous speech corpus consisted of 50 speakers'speech data that free telephone voice.2. Constructed a key-word recognition system based on continuous hidden Markov model framework. The system of construction was involved in speech processing, feature extraction, acoustic modeling and training, keyword detection, keyword recognition and so on.3. The third part is designing and implementation about key-word recognition system. To save the identification time, it uses the matching multi-threading mechanism in longer features the current frame and the code distance calculation part to optimize the recognition module, and added the endpoint detection, using energy and zero-crossing to control, and have been designed recording module to call Windows system function, complete the entry, storage of sound, and feature extraction module converts the audio files into .mfc file, as well as the resulting output waveform display module and display module will be converted into the Latin alphabet Uighur to display.