本文针对语音识别中HMM模型需要大量训练,而在某些实际应用中不可能训练多次的问题,提出一种基于余弦整形变换的变帧率训练方法,并在人名声控拨号系统中进行实验,在训练一次的条件下,系统识别率提高4.2%。实验表明,该方法对解决语音识别系统中训练数据少的问题具有明显效果。
Abstract
In speech recognition HMM requires a large number of data for training , however ,in some applications it is impractical. Therefore ,a VFR training method based on pattern transform method with consine function is presented ,in this paper ,to solve this problem. We apply this original method to a voice control dialling system. System accuracy increases 4.2% on the condition of training just only one time. It isdemonstrated that this method has obvious effect on the scarcity of training data.
关键词
语音识别 /
HMM /
余弦整形变换 /
变帧率
{{custom_keyword}} /
Key words
speech recognition /
HMM /
patternt ransform method with cosine function /
variable frame rate
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Rabiner L R. A tutorial on hidden markov models and selected applications in speech recognition. Proc. IEEE. Feb. 1989 ,77 (2) :257～286
[2] 杨行峻,迟惠生等. 语音信号数字处理. 北京:电子工业出版社,1995
[3] Ponting K M , Peeling S M. The use of variable frame rate analysis in speech recognition. Computer Speech and Language ,1991 ,(5) :169～179
[4] 孙放,胡光锐,虞晓. 变帧率技术在语音识别中的应用. 上海交通大学学报,1998 ,(8)
[5] Guo J ,Sun N ,Nemoto Y et al . Recognition of Handwritten Characters Using Pattern Transformation Method with Cosine Function. Trans. IEICE ,Apr. 1993 ,J76 - D - Ⅱ(4) :835～842
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
高等学校骨干教师资助计划教技司(2000-65)
{{custom_fund}}
PDF(295 KB)
