汉语数码语音识别分析毕业设计论文
Abstract
This paper reports a high-performance MDSR system. In the MDSR system, MFCC (Mel Frequency Cepstrum Coefficient) is used as the main feature parameter of the speech. And the formant trajectory and the nasal feature are extracted to discriminate the confusable words. The feature-based real-time endpoint detection algorithm is proposed to reduce the requirement of system resource and improve the disturbance-proof ability. The two-stage recognition frame is adopted to enhance the discriminating ability, in which the candidate words are decided in the first stage and the confusable word pairs are discriminated in the second stage.
Mandarin digit string speech recognition are studied in this paper. Mandarin digit string speech recognition is of great value in the applications of speech-dialing telephone, remote controlling of consumer devices and industrial controlling, et al. Compared with the English digit speech recognition, the mandarin digit speech recognition is still underdeveloped. The reason is due to that the confusion among the mandarin digits is fairly high and the co-articulation is relatively serious. This paper utilizes the MFCC parameters and context-independent word-based continuous HMM as the acoustic models. The segment duration information is introduced in the form of duration-dependent transition probability. In order to output multiple candidates, the multi-candidate frame-synchronous algorithm is used as the search strategy.
Speaker adaptation is one of the effective approaches to improve the performance of speaker-independent speech recognition. This paper applies the MAP algorithm to the Mandarin digit speech recognition, and discusses several approaches that accelerate the adaptaion process as well as the effects on the unadapted speakers. The experiment shows that the MAP algorithm can considerably reduce the error rate of Mandarin digit speech recognition on the adapted speaker and cause little influence on the unadapted speakers.
Keywords : Mandarin digit (string) speech recognition speaker adaptation
汉语数码语音识别分析毕业设计论文汉语数码语音识别分析毕业设计论文隐藏>> 第三章 识别算法改进 第三章.识别算法改进§3.1 基本识别算法 3.1.1 Viterbi 算法 用...
顾良. 汉语数码语音识别方法研究及 DSP 系统设计:[博士学位论文]. 北京:清华大学 电子工程系,1997 5. 李须真. 汉语数码语音识别在定点 DSP 上的实现:[硕士...
汉语数码语音识别分析毕业设计第四章_工学_高等教育_教育专区。汉语数码语音识别分析毕业设计论文第四章 汉语数码串语音分析 第四章 汉语数码串语音识别分析§4.1...
汉语数码语音识别分析毕业设计论文汉语数码语音识别分析毕业设计论文隐藏>> 第五章 汉语数码语音识别自适应 第五章 汉语数码语音识别自适应§5.1 自适应简介 5.1...
汉语数码语音识别分析毕业设计论文汉语数码语音识别分析毕业设计论文隐藏>> 第六章 总结与展望 第六章 总结与展望 §6.1 本论文工作总结本论文主要介绍了 MDSR ...
第二章 汉语数码语音的特征提取及端点检测方法分析 第二章 汉语数码语音的特征提取 及端点检测方法分析 §2.1 语音特征参数提取语音特 2.1.1 语音特征参数:LPCC...
汉语数码语音识别:发展现状,难点分析与方法比较_经济学_高等教育_教育专区。...摘要】 文简 述了汉语 码语音识别的发展 与现状 ,分 析了汉语数 码语 音...
基于HTK的汉语数码语音识别系统,_专业资料。HTK(是一套专门用于建立和处理删的...[ 摘要] H 是一套专门用于建立和处理H M TK M 的语音处理工具箱,主要应用...
基于HMM的汉语数字语音识别系统研究_信息与通信_工程科技_专业资料。毕业设计...钱盛友 20060401 中文摘要 语音是人与机器之间进行交流的重要工具,语音识别技术就...
基于HTK的连续汉语数码语音输入系统研究_专业资料。...文章在详细介绍了语音识别的过程、原理及相关概念的... 502)505 ( 贵州大学计算机科学与信息学院 摘要 ...
热门文档