摘要
为了提高噪声环境下说话人识别系统的识别性能,将基于听觉掩蔽效应的语音增强技术作为预处理器,对语音信号首先进行降噪处理,提高输入信号的信噪比。实验证明,经过降噪处理的语音信号送入说话人识别系统,提高了系统的识别性能。
Speech enhancement method based on masking properties of the human auditory system is used to reduce the white noise in the front-end. Experimental results show that the performance of speaker recognition system is improved with the processed speech signal in the noise environments.
出处
《电声技术》
2007年第6期41-43,共3页
Audio Engineering
关键词
说话人识别
高斯混合模型
特征提取
掩蔽阈值
语音增强
speaker recognition
gaussian mixture model
feature extract
masking properties
speech enhancement
