稀疏低秩模型及相位谱补偿的语音增强算法
|
摘要
针对现有的语音增强算法存在增强效果差、语音信号失真等问题,提出了稀疏低秩模型及改进型相位谱补偿的语音增强算法。首先,用稀疏低秩模型处理含噪语音的幅度谱,得到分离后的语音。接着,用归一化最小均方自适应滤波算法优化相位谱补偿算法的补偿因子。然后,对稀疏低秩分离后的语音进行改进型相位谱补偿处理,得到最终增强的语音。最后,对增强后的语音进行感知语音质量评价分析及频谱分析。实验结果表明,该方法能够有效地去除噪声,并且在低信噪比的情况下,可以保持语音的清晰度。
Aiming at the problems that the existing speech enhancement algorithms have poor speech enhancement effect and speech signal distortion, a speech enhancement algorithm with sparse low rank model and improved phase spectrum compensation is proposed. First, the sparse low rank model is used to deal with the amplitude spectrum of the noisy speech, and the separated speech is obtained. Second, Normalized Least Mean Square(NLMS)adaptive filtering algorithm is applied to optimize the compensation factor of Phase Spectrum Compensation(PSC)algorithm. In addition,speech after the sparse low rank separation is processed by the improved phase spectrum compensation to obtain the enhanced speech. Finally, the Perceptual Evaluation of Speech Quality(PESQ)and spectrum are used to analyze the enhanced speech. The experimental results show that the noise can be effectively removed with the method, and in the case of low signal to noise ratio, the clarity of speech can be maintained with the method.
Aiming at the problems that the existing speech enhancement algorithms have poor speech enhancement effect and speech signal distortion, a speech enhancement algorithm with sparse low rank model and improved phase spectrum compensation is proposed. First, the sparse low rank model is used to deal with the amplitude spectrum of the noisy speech, and the separated speech is obtained. Second, Normalized Least Mean Square(NLMS)adaptive filtering algorithm is applied to optimize the compensation factor of Phase Spectrum Compensation(PSC)algorithm. In addition,speech after the sparse low rank separation is processed by the improved phase spectrum compensation to obtain the enhanced speech. Finally, the Perceptual Evaluation of Speech Quality(PESQ)and spectrum are used to analyze the enhanced speech. The experimental results show that the noise can be effectively removed with the method, and in the case of low signal to noise ratio, the clarity of speech can be maintained with the method.
引文
[1]Islam M T,Shahnaz C,Fattah S A.Speech enhancement based on a modified spectral subtraction method[C]//IEEE International Midwest Symposium on Circuits and Systems,2014:1085-1088.
[2]Lu C T,Tseng K F,Chen Y Y,et al.Speech enhancement using spectral subtraction algorithm with over-subtraction and reservation factors adapted by harmonic properties[C]//International Conference on Applied System Innovation,2016:1-5.
[3]Prabhakaran G,Indra J,Kasthuri N.Tamil speech enhancement using non-linear spectral subtraction[C]//International Conference on Communications and Signal Processing,2014:1482-1485.
[4]Upadhyay N,Karmakar A.Speech enhancement using spectral subtraction-type algorithms:A comparison and simulation study[J].Procedia Computer Science,2015,54:574-584.
[5]Ganesan V,Manoharan S.Surround noise cancellation and speech enhancement using sub band filtering and spectral subtraction[J].Indian Journal of Science&Technology,2015,8(33).
[6]Biswas T,Mandal S B,Saha D,et al.A novel reconfigurable hardware design for speech enhancement based on multi-band spectral subtraction involving magnitude and phase components[J].Computer Science,2015.
[7]Ykhlef F,Ykhlef H.A smoothed minimum mean-square error log-spectral amplitude estimator for speech enhancement[C]//International Conference on Multimedia Computing and Systems,2014:246-249.
[8]Trawicki M B,Johnson M T.Beta-order minimum meansquare error multichannel spectral amplitude estimation for speech enhancement[J].International Journal of Adaptive Control&Signal Processing,2015,29(10):1287-1295.
[9]Sardaroudi S G,Geravanchizadeh M.A perceptual subspace approach for speech enhancement[C]//International Symposium on Telecommunications,2010:878-881.
[10]Wright J,Ganesh A,Rao S,et al.Robust principal component analysis:Exact recovery of corrupted low-rank matrices[J].Advances in Neural Information Processing Systems,2009,87(4).
[11]Stark A P,Wojcicki K,Lyons J,et al.Noise driven short-time phase spectrum compensation procedure for speech enhancement[J].Proceedings Interspeech,2008.
[12]Sun P,Qin J.Low-rank and sparsity analysis applied to speech enhancement via online estimated dictionary[J].IEEE Signal Processing Letters,2016,23(12):1862-1866.
[13]Lin Z,Chen M,Ma Y.The Augmented lagrange multiplier method for exact recovery of corrupted low-rank matrices[J].Eprint Arxiv,2010,9.
[14]Li Z,Wu W,Zhang Q,et al.Speech enhancement using magnitude and phase spectrum compensation[C]//IEEE/ACIS International Conference on Computer and Information Science,2016:1-4.
[15]Zhang S,Zhang J.Transient analysis of zero attracting NLMS algorithm without Gaussian inputs assumption[J].Signal Processing,2014,97(7):100-109.
[2]Lu C T,Tseng K F,Chen Y Y,et al.Speech enhancement using spectral subtraction algorithm with over-subtraction and reservation factors adapted by harmonic properties[C]//International Conference on Applied System Innovation,2016:1-5.
[3]Prabhakaran G,Indra J,Kasthuri N.Tamil speech enhancement using non-linear spectral subtraction[C]//International Conference on Communications and Signal Processing,2014:1482-1485.
[4]Upadhyay N,Karmakar A.Speech enhancement using spectral subtraction-type algorithms:A comparison and simulation study[J].Procedia Computer Science,2015,54:574-584.
[5]Ganesan V,Manoharan S.Surround noise cancellation and speech enhancement using sub band filtering and spectral subtraction[J].Indian Journal of Science&Technology,2015,8(33).
[6]Biswas T,Mandal S B,Saha D,et al.A novel reconfigurable hardware design for speech enhancement based on multi-band spectral subtraction involving magnitude and phase components[J].Computer Science,2015.
[7]Ykhlef F,Ykhlef H.A smoothed minimum mean-square error log-spectral amplitude estimator for speech enhancement[C]//International Conference on Multimedia Computing and Systems,2014:246-249.
[8]Trawicki M B,Johnson M T.Beta-order minimum meansquare error multichannel spectral amplitude estimation for speech enhancement[J].International Journal of Adaptive Control&Signal Processing,2015,29(10):1287-1295.
[9]Sardaroudi S G,Geravanchizadeh M.A perceptual subspace approach for speech enhancement[C]//International Symposium on Telecommunications,2010:878-881.
[10]Wright J,Ganesh A,Rao S,et al.Robust principal component analysis:Exact recovery of corrupted low-rank matrices[J].Advances in Neural Information Processing Systems,2009,87(4).
[11]Stark A P,Wojcicki K,Lyons J,et al.Noise driven short-time phase spectrum compensation procedure for speech enhancement[J].Proceedings Interspeech,2008.
[12]Sun P,Qin J.Low-rank and sparsity analysis applied to speech enhancement via online estimated dictionary[J].IEEE Signal Processing Letters,2016,23(12):1862-1866.
[13]Lin Z,Chen M,Ma Y.The Augmented lagrange multiplier method for exact recovery of corrupted low-rank matrices[J].Eprint Arxiv,2010,9.
[14]Li Z,Wu W,Zhang Q,et al.Speech enhancement using magnitude and phase spectrum compensation[C]//IEEE/ACIS International Conference on Computer and Information Science,2016:1-4.
[15]Zhang S,Zhang J.Transient analysis of zero attracting NLMS algorithm without Gaussian inputs assumption[J].Signal Processing,2014,97(7):100-109.