摘要
为实时消除现场录音过程中的不相关平稳加性噪声,采用基于先验信噪比的谱减法语音去噪算法,对其增益因子进行了修正,提出一种简化的语音激活检测(Voice Activity Detection,VAD)算法用于噪声参数更新。总结出4条DSP实现时的优化措施,使算法运算量大大降低,方便算法的实时实现。用本方法在TMS320C6713浮点DSP上实现了两通道的语音去噪并用于一会议室的录音系统中,取得良好的噪声消除效果。
To enhance speech degraded by uncorrelated additive noise of recording system in real time,spectral subtraction algorithm based on prior SNR Estimation was used.The gain factor of spectral subtraction algorithm was improved.To ensure signal processing in real time,a simple VAD algorithm and noise parameter updating method were proposed and 4 optimizing measures were summarized.The computational complexity was greatly reduced.Two channels noise reduction was realized on TMS320C6713.The application of the product in the recording system of a conference room proved that the noise reduction effect was good.
出处
《太原理工大学学报》
CAS
北大核心
2010年第6期700-704,共5页
Journal of Taiyuan University of Technology
基金
广东省科技计划资助项目(2008B011400001)