摘要
为了改善谱减法存在的“音乐噪声”残余大,重构语音时使用带噪相位的缺陷,提出了一种联合相位谱补偿的改进型谱减法语音增强算法。引入改进型噪声估计方法来得到更准确的噪声谱,依据带噪语音谱与估计的噪声谱的信噪比构造一个线性函数来调整过减因子进行谱减,通过相位补偿函数对带噪语音的相位谱进行补偿,将谱减后的语音幅度谱与补偿后的相位谱进行结合,通过IFFT及重构得到去噪语音。仿真实验表明,在使用联合相位谱补偿的改进型谱减法对带噪语音进行处理后,语谱图更加清晰,PESQ及STOI评分也有所提高。
In order to improve the two defects of spectral subtraction,large"music noise"residue and the use of noisy phase in speech reconstruction,the improved spectral subtraction speech enhancement algorithm with joint phase spectrum compensation is proposed.Firstly,an improved noise estimation method is introduced to obtain a more accurate noise spectrum,and then a linear function is constructed according to the signal-to-noise ratio of noisy speech spectrum and estimated noise spectrum to adjust the over subtraction factor for spectral subtraction.Then the phase spectrum of noisy speech is compensated by phase compensation function.Finally,the speech amplitude spectrum after spectrum subtraction is combined with the compensated phase spectrum,and the denoised speech is obtained by IFFT and reconstruction.The experimental simulation results show that after using the improved spectral subtraction with joint phase spectrum compensation to process noisy speech,the spectrogram is clearer,and the scores of PESQ and STOI are also improved.
作者
刘龙威
夏秀渝
LIU Longwei;XIA Xiuyu(College of Electronic and Information Engineering,Sichuan University,Chengdu 610065,China)
出处
《计算机与网络》
2022年第13期69-73,共5页
Computer & Network
关键词
噪声估计
相位谱补偿
谱减法
语音增强
noise estimation
phase spectrumcompensation
spectral subtraction
speech enhancement
