基于HTK的白族语音识别方法

J4 ›› 2013, Vol. 12 ›› Issue (10): 27-31.

基于HTK的白族语音识别方法

大理学院工程学院，云南大理 671003

收稿日期:2013-03-16 修回日期:2013-08-16 出版日期:2013-10-15 发布日期:2013-10-15
作者简介:张令通，副教授，主要从事语音信号处理及通信与信息系统研究.
基金资助:
云南省教育厅科学基金资助项目（2012Y154）

The Speech Recognition Method of the Bai's Language Based on HTK

College of Engineering，Dali University, Dali, Yunnan 671003,China

Received:2013-03-16 Revised:2013-08-16 Online:2013-10-15 Published:2013-10-15

摘要/Abstract

摘要：

利用计算机识别少数民族语音是保护和传承民族文化的重要手段。白族是祖国西南边陲重要的少数民族之一，其历
史悠久，文化灿烂。为实现使用白族语进行人与计算机的语音交互，提出了一种基于HTK的白族语音词识别方法。该方法针对白族语的发音特点，以音素为基本识别单元，利用HTK工具提取39维MFCC语音特征参数，构建HMM模型，采用Viterbi算法进行模型训练和匹配来实现白族语音的识别。实验表明，算法的识别准确率达到93.3%。该方法识别准确率高，为研究少数民族语音识别提供了有益的借鉴。

Abstract:

Using computer to identify the speech sounds of minority nationality is an important means to protect and inherit the national culture. The Bai is one of the important minority nationalities in the southwest border of the motherland. It has a long history and the splendid culture. In order to implement human-computer interaction in Bai' s language, a sound recognition method for Bai’s language based on HTK is proposed. According to characteristics of Bai's language, this method structures the HMM model through using phonemes as the basic units and using the HTK tool to extract the speech feature parameters of 39-dimensional MFCC. Viterbi algorithm is used to train the model and match to achieve speech recognition of Bai's language. The experimental result shows that the recognition system can obtain a high recognition rate（up to 93.3%）and also provides a useful reference for the speech recognition research
on minorities' languages.

中图分类号:

TP391

张令通. 基于HTK的白族语音识别方法[J]. J4, 2013, 12(10): 27-31.

ZHANG Ling-Tong. The Speech Recognition Method of the Bai's Language Based on HTK[J]. J4, 2013, 12(10): 27-31.

参考文献

〔1〕赵金灿, 闫正锐, 张钰芳.白族语言使用现状及语言态度调查〔J〕.大理学院学报,2012,11(8):31-35.
〔2〕徐琳,赵衍荪.白语简志〔M〕.北京：民族出版社,1984.
〔3〕Lipeika Antanas,Lipeikiene Joana. On the use of the formant features in the dynamic time warping based recognition of isolated words〔J〕. Informatica,2008,19(2):213-226.
〔4〕Chaiwongsai, Jirabhorn. An architecture of HMM-based isolated-word speech recognition with tone detection function〔C〕/ 2008 International Symposium on Intelligent Signal Processing and Communication Systems. ISPACS, 2008.
〔5〕魏巍,张海涛. 一种基于HTK的数字语音识别系统〔J〕.计算机系统应用,2011,20(9):17-21.
〔6〕王炳锡, 屈丹, 彭煊. 实用语音识别基础〔M〕. 北京: 国防工业出版社, 2005.
〔7〕涂俊辉,续晋华.基于HTK的连续语音识别系统及其在TIMIT上的实验〔J〕.现代计算机,2009,319(11):29-33.
〔8〕Yuan Lichi. An improved HMM speech recognition model〔C〕/ 2008 International Conference on Audio, Language and Image Processing. 2008:1311-1315.
〔9〕杨建华,于小宁.说话人识别中语音特征参数研究〔J〕.大理学院学报,2009,8(8):32-35.
〔10〕Fujimura H. N-Best rescoring by adaboost phoneme classifiers for isolated word recognition〔C〕/ 2011 IEEE
Workshop on Automatic Speech Recognition & Understanding (ASRU).2011:83-85.
〔11〕Im Jung-Hui, Lee Soo-Young. Unified Training of Feature Extractor and HMM Classifier for Speech Recognition
〔J〕. Signal Processing Letters,2012,19(2):111-114.
〔12〕曾妮,费洪晓,姜振飞.基于HTK的特定词语音识别系统〔J〕.计算机系统应用,2011,20(3):157-160.
〔13〕Kazemi A R. Isolated word recognition based on intelligent segmentation by using hybrid HTD- HMM〔C〕/ 5th
WSEAS International Conference on Circuits, Systems, Signal and Telecommunications (CISST' 11) . 2011: 38-41.