论文部分内容阅读
为了在噪声抑制和语音失真中之间寻找最佳平衡,提出了一种听觉频域掩蔽效应的自适应β阶贝叶斯感知估计语音增强算法,以期提高语音增强的综合性能。算法利用了人耳的听觉掩蔽效应,根据计算得到的频域掩蔽阈自适应调整β阶贝叶斯感知估计语音增强算法中的β值,从而仅将噪声抑制在掩蔽阈之下,保留较多的语音信息,降低语音失真。并分别用客观和主观评价方式,对所提出的算法的性能进行了评估,并与原来基于信噪比的自适应β阶贝叶斯感知估计语音增强算法进行了比较。结果表明,频域掩蔽的β阶贝叶斯感知估计方法的综合客观评价结果在信噪比为-10 dB至5 dB之间时均高于基于信噪比的自适应β阶贝叶斯感知估计语音增强算法。主观评价结果也表明频域掩蔽的β阶贝叶斯感知估计方法能在尽量保留语音信息的同时,较好的抑制背景噪声。
In order to find the best balance between noise suppression and speech distortion, a speech enhancement algorithm based on adaptive β-Bayesian perceptive estimation of audio frequency domain masking effect is proposed in order to improve the overall performance of speech enhancement. Based on the calculated masking threshold of frequency domain, the algorithm adaptively adjusts the β value in the speech enhancement algorithm of β-order Bayesian perception to suppress the noise only under the masking threshold and retain more Voice message, reduce speech distortion. The performances of the proposed algorithm are evaluated objectively and subjectively, respectively, and compared with the original speech enhancement algorithm based on adaptive β-order Bayesian estimation. The results show that the comprehensive objective evaluation results of β-order Bayesian estimation in frequency domain masking are both higher than SNR-based adaptive β-order Bayesian perception when SNR is between -10 dB and 5 dB Estimation of speech enhancement algorithms. The subjective evaluation results also show that the method of β-order Bayesian estimation in frequency domain masking can keep the voice information as much as possible and suppress the background noise better.