摘要
混响是声音经过室内墙壁等物体反射、吸收后多径传播叠加产生的,是导致语音识别系统性能下降的主要因素之一。基于TF-GSC的混响消除算法在估计混响功率谱时可能会出现过估计的现象,导致输出语音失真。提出一种直达声补偿策略,并将其应用到混响抑制算法中去。实验结果表明,直达声补偿策略减小了输出语音失真,提高了输出语音质量。
The reverberation is generated when the speech signal is reflected and absorbed by wall and other objects, and is superimposed by multipath propagation, which is one of the main reasons that degrade the performance of speech recognition systems. The dereverberation algorithm based on TF-GSC may lead to excessive estimation of reverberation power spectrum, resulting in the distortion of output. In this paper, a direct path compensation strategy is proposed and applied to the dereverberation algorithm. The experimental results show that the direct path compensation strategy reduces the output speech distortion and improves the quality of output speech.
作者
陈长海
Chen Changhai(Fuzhou RockChip Electronics Co. , Ltd. , Fuzhou 350003, Chin)
出处
《微型机与应用》
2017年第24期32-36,共5页
Microcomputer & Its Applications
关键词
混响
TF-GSC
直达声补偿
混响功率谱估计
reverberation
TF-GSC
direct path compensation
reverberation power spectrum
