摘要
论文针对强噪音环境提出一种PZT振动拾音器作为语音识别的前端输入。在同步采样前提下,比较了其信号与麦克风信号在时、频域上抗噪声性能上的差异。为弥补部分语音信息的丢失,提出了一种混合倒谱系数作为语音识别的特征,并分析了其抗噪声性能。
A noise-free PZT sensor is made as speech signal input equipment to an auto speech recognition system.Its noise restraint characteristics are compared with that of normal microphones.The results show that the sensor is much better to suppress strong background noise than normal microphone.To the sensor,Compound Cepreum Coefficients(CD-CC)is proposed to substitute the Linear Predication Cepreum Coefficients(LPCC)in the speech recognition system.Experi-ments prove that CDCC is prior to LPCC in both stability and noise restraint.
出处
《计算机工程与应用》
CSCD
北大核心
2003年第27期113-117,共5页
Computer Engineering and Applications
