摘要
随着移动通信技术的快速发展,语音增强的研究及其实际应用成为数字化通信的一个重要的研究方向。在数字信号处理技术的支撑下,许多优秀的语音增强算法的实时实现成为了可能。谱减法是一种运算量相对较小,增强效果明显,并且容易实时实现的语音增强算法,但是其缺点就是残留有音乐噪声。针对传统谱减法,本语音增强系统采用了一种改进算法,就是按bark频率划分频带后在各个子带上与噪声估计相减,并后置残余噪声衰减算法,测试表明,带噪语音信号经过该系统的增强后,语音质量得到了明显改善。
With the rapid development of mobile communication technology,the research on speech enhancement and its practical applications has become an important research direction of digital communication.The real-time implementation seems to be possible in support of digital signal processing technology.Spectral subtraction is a speech enhancement algorithm,integrating the advantages of relatively small operation,obvious enhanced effect and easy real-time implementation,but the disadvantage is that it has residual musical noise.In connection with the traditional spectral subtraction,this speech enhancement system uses a modified algorithm,which uses bark frequency to divide the band and subtracts with noise estimation in each sub-band and last is the residual noise reduction algorithm.Test showed that the noisy speech signal is enhanced through the system and voice quality has been significantly improved.
出处
《电子测试》
2011年第10期46-50,共5页
Electronic Test
关键词
语音增强
谱减法
DSP
speech enhancement
spectral subtraction
DSP
