摘要
基音检测是河南方言语音信号处理中的一个重要环节,针对低信噪比环境下的河南方言语音基音检测准确率低的问题,提出了一种语音信号增强和基音检测相结合的算法。通过多窗谱估计的改进谱减法对语音信号进行降噪处理,对增强后的语音信号用中心削波法消除偏离基音轨迹的野点,再通过自相关法实现基音检测。仿真结果表明,对于低信噪比环境下河南方言语音信号的基音估值检测结果准确,估算出的基音频率和实际基音频率能很好的重合。
Pitch detection is an important part of Henan dialect speech signal processing, for low accuracy issues of Henan dialect pitch detection of noisy speech, the improved pitch detection method was proposed for noisy speech combined with speech enhancement. It reduces speech signal noise by using modified spectral subtraction for multiple window spectral, centre clipping method eliminates outliers deviate from the pitch track for enhanced speech signal, realized pitch detection by ACF. Experimental simulations results show that, the method can detect Henan dialect pitch detection in low SNR, and the estimated pitch frequency coincides with the actual pitch frequency.
出处
《电声技术》
2016年第4期51-54,共4页
Audio Engineering
基金
郑州市科技局科技发展计划项目(20140663)
郑州市嵌入式系统应用技术重点实验室建设项目(121PYFZX177)
关键词
基音检测
多窗谱
谱减法
自相关函数
pitch detection
multiple window spectrum
spectral subtraction
ACF
作者简介
韩芳（1981-），女，硕士，副教授，主要研究方向为语音信号处理、计算机网络。
