The research on Uighur speaker-dependent isolated word speech recognition

The research on Uighur speaker-dependent isolated word speech recognition Wushour·silamu Caiqin Nuominghua College of information science and engineering Xinjiang University, Urumqi 830046 Abstract: A small vocabulary, isolated word speech recognition system in java has been realized. In this system we have done the extraction of feature parameter, the training of speech model parameter and the recognition of the recorded speech. MFCC is used as feature parameter, HMM model is used for speech model in this article and do some study on Uighur speech recognition based on them. Keywords：Speech recognition. Hidden Markov models. Speech feature parameter. Ⅰ.Introduction Based on achieving English speech recognition system ,we study in Uighur language and Chinese language voice recognition system. We using Jbuilder and Jdk 1.4 development tools to study how to deal with the Uighur text characters. Isolated word speech recognition system, with broad application prospects, such as computer control, industrial control orders, family services, banking services, personal information Identification, personal items such as mobile communications, the application of this technology will greatly facilitate the daily lives of people. The work of Uighur language recognition system further study instructive and help. Ⅱ. Discussion Uighur language speech recognition in the development process to address issues In system development process, I base on Chinese speech recognition system transform into Uighur language speech recognition system, which is facing the biggest problem how to deal with Uighur text characters in applications processing, Uighur language of a different editorial direction, wide-ranging characters, ranging from long code, even set spacing, and other characteristics of the word, all these specific issues to be resolved. In java programming, each character stored in the code , Internal code is Unicode (all languages are corresponding codes, internal computers character string expressed internal code, in general computer character string encoding relates platform, the Java unrelated platform using Unicode). When Java get a character string form a byte stream, associated with the platform byte changes Unicode byte character string which independent on platform. Exporting, Java will be changes Unicode character string into the byte stream which relates with platform. if a Unicode characters do not exist in a given platform will be exporting a '? '. 360 1, Java Translation devices translate the source document, source document first will be converted to Unicode coding. So before translation of documents, we must "tell" Translation devices what coding method used. For example : Our source document is preserved in UTF-8 method, in translation, but translation devices regard as preserving GBK method .the Translation devices will converted accordance with GBK to Unicode coding method of source documents, and then translation, so certainly wrong, Indeed Translation devices should follow the converted accordance with UTF-8 to Unicode coding method. For control panels procedures, translation device will source documents as system acquiescence coding type to coding (the coding system acquiescence depend on control panels configuration, Chinese Win2K type is usually GBK), can use -encoding parameters to configure, such as : Javac -encoding UTF-8, such translation device will be used as the source document …

Content truncated to 500 words