基于Walsh变换的汉语非特定人语音识别的研究  

The Study of Speaker-independent Chinese Digit Recognition Based on Walsh Transform

作　　者：田岚[1] 白树忠 董春阳[2] 

机构地区：[1]山东工业大学,济南250061 [2]济南市电信局

出　　处：《控制与决策》1997年第2期142-145,共4页Control and Decision

摘　　要：运用Walsh变换对语音信号进行分析,得到了语音信号的Walsh域特征——2D自由度功率谱特征。利用此特征采用神经元网络法对汉语数字音0～9进行不定人语音识别,结果表明,该特征不仅能给出足够的识别信息,而且由于Walsh变换只有加/减运算,因而在识别速度方面比常用的频域特征有明显提高。Based on the analysis of speech signal by use of Walsh transform, a new characteristic parameter -the 2 D freedom power spectrum is obtained and used in a Chinese digit speaker - independent speech recong-nition system which uses neural networks. The results show that this character can not only provide enough information about speech signal, but also provide a faster response speed of recongnition system, because the character is extracted only by using add/substract operation in Walsh transform, no multiply operation.

关 键 词：语音识别 神经元网络 WALSH变换 

分 类 号：TN912.34[电子电信—通信与信息系统]