摘要
介绍一个连续语音识别系统的构成以及HTK工具的使用,然后尝试利用该工具搭建一个连续语音识别系统,使用该系统在一个非特定人的大词汇量连续语音数据库——TIMIT上进行实验,讨论如何选择不同的声学单元进行建模,并且对模型的参数进行一系列的改进。
As the society advances, there has been an urgent need for automatic speech recognition technologies, especially the continuous speech recognition which has significant applications in different fields. HTK is short for h idden markov model toolkit. It＇s an toolbox based on HMM, originally developed by Cambridge University in C language. It has now been widely used in speech recognition and speech synthesis. Introduces an approach for constructing a continuous speech recognition system with HTK, and gives some experiments on the TIMIT database with this recognizer, discusses how to select differnet acoustic units for modeling, and makes some improvements to the parameters of the model.
出处
《现代计算机》
2009年第11期29-33,共5页
Modern Computer
关键词
连续语音识别
隐马尔可夫模型
HTK
TIMIT
Continuous Speech Recognition
Hidden Markov Model
Hidden Markov Model Toolkit
TIMIT
