重庆方言语音识别系统的设计与实现

Design and Implementation of Speech Recognition System in Chongqing Dialect

摘要:

语音识别赋予了计算机能够识别出语音内容的功能，是人机交互技术领域的重要研究内容。随着计算机技术的发展，语音识别已经得到了成熟的发展。但是关于方言的语音识别还有很大的发展空间。中国是一个幅员辽阔、人口众多的国家，因此方言种类繁多，其中有3000多万人交流使用的重庆方言就是其中之一。采集了重庆方言的部分词语的文本文件和对应的语音文件建立语料库，根据重庆方言的发音特点，选取重庆方言的声韵母作为声学建模基元，选取隐马尔可夫模型(Hidden Markov Model, HMM)为声学模型设计了一个基于HMM的重庆方言语音识别系统。在训练过程利用语料库中训练集语料对声学模型进行训练，形成HMM模型库；在识别过程利用语料库中的测试集语料进行识别测试。实验结果表明，该系统能够实现重庆方言的语音识别，并且识别的正确率为100%。

Abstract:

Speech recognition has given the computer the ability to recognize the function of speech content, and it is an important research content in the field of human-computer interaction technology. With the development of computer technology, speech recognition has been developed very well. But there is still a lot of room for speech recognition in dialects. China is a country with a vast territory and a large population, so there are many dialects, among which more than 30 million people communicate with Chongqing dialect is one of them. The text files and corresponding speech files of some words in Chongqing dialect are collected to establish corpus. According to the pronunciation characteristics of Chongqing dialect, the vowel of Chongqing dialect is selected as the recognition primitive. HMM model for acoustic model to design Chongqing dialect speech recognition system. In the training process, the training corpus is used to train the acoustic model, and then the HMM model library is formed. In the recognition process, the test corpus is used to test the corpus. The experimental results show that the designed system can realize the speech recognition of Chongqing dialect, and the recognition rate is 100%.