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.
出处
《控制与决策》
EI
CSCD
北大核心
1997年第2期142-145,共4页
Control and Decision
