语音通信降噪研究

语音通信系统中，语音通过信道传输将不可避免地引入码间串扰和信号畸变，同时受到噪声污染。本文在分析自适应盲均衡算法CMA(constant modulus algorithm）和改进盲均衡算法的基础上，考虑到自适应盲均衡技术在语音噪声控制方面能力有限，将自适应盲均衡技术与小波包掩蔽阈值降噪算法联合使用，形成一种基带语音增强新方法。仿真试验结果显示自适应盲均衡技术可以使星座图变得清晰而紧凑，有效减小误码率。研究证实该方法在语音信号ISI和畸变严重情况下，在白噪及有色噪声不同的噪声环境中都具有稳定的降噪能力，消噪同时可获得汉语普通话良好的听觉效果。     

采用联合字典优化的噪声鲁棒性语音转换算法

针对含噪语音难以实现有效的语音转换,本文提出了一种采用联合字典优化的噪声鲁棒性语音转换算法。在联合字典的构成中,语音字典采用后向剔除算法(Backward Elimination algorithm,BE)进行优化,同时引入噪声字典,使得含噪语音与联合字典相匹配。实验结果表明,在保证转换效果的前提下,后向剔除算法能够减少字典帧数,降低计算量。在低信噪比和多种噪声环境下,本文算法与传统NMF算法和基于谱减法消噪的NMF转换算法相比具有更好的转换效果,噪声字典的引入提升了语音转换系统的噪声鲁棒性。      

一种改进的EMD硬阈值去噪算法

为了提升基于经典小波阈值的EMD去噪算法的性能,利用高斯白噪声的统计特征提出了一种改进的硬阈值去噪算法;首先将含噪信号进行EMD分解,把第一个固有模态函数作为高频噪声直接去除并估算出其他 IMF中高斯白噪声的能量,然后根据硬阈值去噪的原理,利用滤除掉的样本点包含的能量等于白噪声的能量确定出合适的阈值;该方法能根据样本点自适应地确定阈值;最后通过对含噪正弦信号和仿真心电信号的去噪实验证实了改进后的阈值使算法去噪效果有明显提升。     

二次谐波的变分模态分解和小波阈值函数降噪

针对利用可调谐半导体激光器吸收光谱学(TDLAS)技术测量气体浓度过程中二次谐波谱线存在的外界噪声干扰问题,提出一种基于变分模态分解和小波阈值函数复合算法的二次谐波降噪方法。首先对二次谐波含噪信号进行分解,得到有用固有模态函数(IMF)并进行重构,再对重构信号进行小波阈值函数降噪处理。讨论了变分模态分解中最佳平衡参数的选取,得出最佳平衡参数与含噪信号中噪声成正比的结论。通过改变小波变换的阈值函数改变高频小波系数,以更好地抑制噪声。对实际测量曲线的降噪结果表明,所提出的降噪方法可以在信噪比较低的情况下有效抑制噪声,提取有用的二次谐波信号。     

语音识别预处理中的ICA噪声适应性分析

针对高噪环境下语音识别的困难,提出一种基于独立分量分析的盲分离(ICA/BSS)与小波联合的语音降噪预处理方法,针对不同种类和不同输入信噪比的噪声设计了试验,结果表明基于ICA的语音识别预处理方法对低输入信噪比情况下多种噪声具有很强的鲁棒性和优越性,此结论对现实世界高噪环境下的信号分析和语音识别具有重要意义。     

传声器阵列波束比判决语音增强方法

针对单一波束形成器难以深度抑制空间相干干扰的问题,提出了一种综合了最小方差无畸变响应波束形成器与对称子阵延时求和波束形成器的语音增强方法。定义了一种波束输出比因子,根据该因子在目标声区域和干扰声区域的幅值变化,给出了采样协方差矩阵对角加载量的调整方法,并进一步利用该因子在后滤波环节对空间干扰进行判决滤波。文中对判决滤波时的上限阈值和下限阈值的实时更新方法给出了说明。所提出的算法能进一步抑制空间干扰和噪声,且可满足实时需要。在传声器圆阵上的实验表明,该方法在输出信干噪比及语音质量上,均优于经典对角加载算法及采样协方差矩阵扫描重构算法。      

扩散噪声下协方差矩阵重构的语音分离与降噪

针对传统多通道语音分离算法在扩散噪声下性能下降的问题,提出了一种用于语音分离及降噪的空间协方差模型及参数估计方法。该方法将扩散噪声视为独立声源,利用由导向矢量重构的空间协方差矩阵建模目标声源的空间特性,并通过空间协方差分析方法估计用于语音分离的多通道维纳滤波器。同时,还提出了一种联合该方法的后置滤波器参数框架,为输出信号降噪和失真的折中提供了更多选择。在扩散噪声下的单目标和多目标实验中,所提方法的语音提取和分离性能都优于对比算法,联合参数的后置滤波器可提供更为符合人们要求的降噪语音,验证了所提模型与参数估计方法的有效性。      

滑动窗累积量的递推估计算法及其在语音端点检测中的应用

提出了一种滑动窗累积量的递推估计算法并应用于语音端点检测中,用以解决传统端点检测方法在噪声环境下检测性能变差的问题。在对含噪语音信号进行加窗之后,利用滑动窗累积量的递推估计算法估计含噪语音信号的高阶累积量值,并在此基础上结合能量特征进行语音端点检测。实验结果表明,所提滑动窗累积量递推估计算法相比较传统高阶累积量计算方法运算效率明显提高;所提端点检测算法在不同噪声和信噪比环境下相比较G.729b算法点正确率Pc-point值平均提升了6.07%。基于滑动窗高阶累积量的语音端点检测算法具有较高的运算效率及良好的鲁棒性。     

滑动窗累积量的递推估计算法及其在语音端点检测中的应用

提出了一种滑动窗累积量的递推估计算法并应用于语音端点检测中,用以解决传统端点检测方法在噪声环境下检测性能变差的问题。在对含噪语音信号进行加窗之后,利用滑动窗累积量的递推估计算法估计含噪语音信号的高阶累积量值,并在此基础上结合能量特征进行语音端点检测。实验结果表明,所提滑动窗累积量递推估计算法相比较传统高阶累积量计算方法运算效率明显提高;所提端点检测算法在不同噪声和信噪比环境下相比较G.729b算法点正确率Pc-point值平均提升了6.07%。基于滑动窗高阶累积量的语音端点检测算法具有较高的运算效率及良好的鲁棒性。      

两阶段复数谱卷积循环网络立体声回声消除

提出了一种两阶段复数谱卷积循环网络(CRN)的立体声回声消除(SAEC)算法,该算法无需对立体声信号进行去相关,因而能够在保证立体声音质和空间感的同时,解决自适应滤波SAEC算法非唯一解问题。所提算法采用两个阶段进行回声消除,第一阶段根据传声器接收信号和参考信号估计回声信号,第二阶段将估计回声信号作为先验信息,联合传声器接收信号作为输入特征,估计近端语音。相对于单阶段CRN算法,该方法能够提高网络对回声和近端语音的区分度,有助于近端语音的提取。另外,网络的输入特征和训练目标均采用复数谱,降低了近端语音的相位估计误差,因而可以进一步提升算法性能。实验表明,基于两阶段复数谱CRN的SAEC算法在单端讲话时的回声抑制量和双端讲话时的语音质量都明显优于传统算法以及单阶段CRN算法。      

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

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。      

一种抑制方向性噪声的双耳语音增强算法

为了给双耳听力设备佩戴者带来更好的语音可懂度,提出了一种利用双耳时间差与声级差的近场语音增强算法,该方法首先利用这两种差异来估计语音的功率谱和语音的相干函数,然后计算干扰噪声在左右耳间的头相关传输函数的比值,最后构造两个维纳滤波器。客观评价的参数显示该算法去噪效果优于对比算法而目标语音的时间差误差和声级差误差低于对比算法。主观的言语接受阈测试表明该方法能有效提高语音可懂度。结果表明,该算法在能够有效去除干扰噪声的同时,保留了目标语音的空间信息。      

小波包自适应阈值语音降噪新算法

为了克服低信噪比输入下,语音增强造成语音清音中的弱分量损失,造成重构信号包络失真的问题。论文提出了一种新的语音增强方法。该方法根据语音感知模型,采用不完全小波包分解拟合语音临界频带,并对语音按子带能量进行清浊音区分处理,在阈值计算上,提出了一种清浊音分离,基于子带信号能量的小波包自适应阈值算法。通过仿真实验,客观评测和听音测试表明,该算法在低信噪比输入时较传统算法,能够更加有效地减少重构信号包络失真,在不损伤语音清晰度和自然度的前提下,使输出信噪比明显提高。将该算法与能量谱减法结合,进行二次增强能进一步提高降噪输出的语音质量。     

基于统计特征和显著度的光谱信号提取算法

各类光谱信号都会受到噪声和基线畸变的影响,在提取光谱信号过程中若不考虑基线畸变和噪声的影响,将会严重影响信号提取的精度和准确性,所以需要在信号提取前消除噪声和基线畸变的影响。大多数信号提取算法的步骤是先提取整体基线,再提取信号,这样难以保证基线的提取精度。为了降低信号提取过程中背景噪声、基线畸变等不利因素的影响,根据信号的存在总是会导致该区域的统计特征不同于背景的特点,提出了一种基于显著度和统计特征的光谱信号检测与提取算法（SSD算法）。首先,在待测数据的不同尺度空间中计算出信号在各尺度下的显著度,将检测出的显著信号点作为候选信号点;其次,利用信号特征去除候选信号点中的伪信号点;最后,对候选信号点所在区域采用二次多项式进行基线拟合以剔除伪信号区域并实现最终的信号提取。为验证SSD算法的综合性能,首先,通过仿真的方法对高斯信号和矩形信号在不同基线类型、不同信噪比下进行实验;然后将该算法与AirPLS算法、Wavelet算法以及DoG算法对两种信号在不同信噪比,不同基线类型下的提取结果进行比较。仿真实验结果表明：与其他算法相比,SSD算法信号提取结果基本不受信号类型和基线畸变类型的影响,且当信噪比大于40时基本不受信噪比的影响;在不同基线畸变类型下,SSD算法对两种信号提取结果的准确度、稳定性、离散度均较好,其他算法则只适用于某种基线畸变类型。从总体提取结果上看,SSD算法提取结果的绝对误差的均值仅为AirPLS算法的8.71%、Wavelet算法的3.52%、DoG算法的2.01%;绝对误差的均方根也仅为AirPLS算法的13.08%、Wavelet算法的5.45%、DoG算法的3.11%。因此,所提出的SSD算法在提取信号时具有良好的综合性能,能够在不同的信噪比及基线畸变情况下准确的提取出信号。     

A comparative intelligibility study of single-microphone noise reduction algorithms

The evaluation of intelligibility of noise reduction algorithms is reported. IEEE sentences and consonants were corrupted by four types of noise including babble, car, street and train at two signal-to-noise ratio levels (0 and 5 dB), and then processed by eight speech enhancement methods encompassing four classes of algorithms: spectral subtractive, sub-space, statistical model based and Wiener-type algorithms. The enhanced speech was presented to normal-hearing listeners for identification. With the exception of a single noise condition, no algorithm produced significant improvements in speech intelligibility. Information transmission analysis of the consonant confusion matrices indicated that no algorithm improved significantly the place feature score, significantly, which is critically important for speech recognition. The algorithms which were found in previous studies to perform the best in terms of overall quality, were not the same algorithms that performed the best in terms of speech intelligibility. The subspace algorithm, for instance, was previously found to perform the worst in terms of overall quality, but performed well in the present study in terms of preserving speech intelligibility. Overall, the analysis of consonant confusion matrices suggests that in order for noise reduction algorithms to improve speech intelligibility, they need to improve the place and manner feature scores.     

Gain-induced speech distortions and the absence of intelligibility benefit with existing noise-reduction algorithms

Most noise-reduction algorithms used in hearing aids apply a gain to the noisy envelopes to reduce noise interference. The present study assesses the impact of two types of speech distortion introduced by noise-suppressive gain functions: amplification distortion occurring when the amplitude of the target signal is over-estimated, and attenuation distortion occurring when the target amplitude is under-estimated. Sentences corrupted by steady noise and competing talker were processed through a noise-reduction algorithm and synthesized to contain either amplification distortion, attenuation distortion or both. The attenuation distortion was found to have a minimal effect on speech intelligibility. In fact, substantial improvements (>80 percentage points) in intelligibility, relative to noise-corrupted speech, were obtained when the processed sentences contained only attenuation distortion. When the amplification distortion was limited to be smaller than 6 dB, performance was nearly unaffected in the steady-noise conditions, but was severely degraded in the competing-talker conditions. Overall, the present data suggest that one reason that existing algorithms do not improve speech intelligibility is because they allow amplification distortions in excess of 6 dB. These distortions are shown in this study to be always associated with masker-dominated envelopes and should thus be eliminated.     

时频结合自适应阈值小波包消噪算法

在充分考虑人耳听觉特性和噪声统计特性的基础上,提出一种时频结合Bark尺度自适应阈值的语音消噪算法,在Bark频域上自适应调整增强系数可以较准确地进行阈值判定。仿真实验验证,时频结合算法在低信噪比输入情况下较传统语音降噪方法具有明显优势,其在消除高斯白噪声的同时有效降低了语音损失,可获得最大信噪比,谱失真测度最小,增强语音的MOS(Mean Opinion Score)评分明显提高,具有较好的听觉效果。     

A wavelet-based forward BSS algorithm for acoustic noise reduction and speech enhancement

In this paper, we address the problem of noise reduction and speech enhancement by adaptive filtering algorithm. Recently, the well known forward blind source separation (FBSS) structure has been largely studied and intensively used to reduce acoustic noise components and to enhance speech signal. The FBSS structure is often combined with adaptive algorithms to accelerate the adaptation of the cross-filters, and to improve noise suppression at the output. In this paper, we propose to use a wavelet transform decomposition in the FBSS structure by using a two-channel forward wavelet symmetric adaptive decorrelating (WFSAD) algorithm. The proposed WFSAD algorithm provides a better compromise between time and frequency resolution and improves robustness of the noise reduction process when compared with the classical two-channel forward symmetric adaptive decorrelating (FSAD) algorithm. Simulation results prove the efficiency of the proposed WFBSS algorithm in comparison with conventional ones in terms of several objective and subjective criteria.     

Effects of directional microphone and adaptive multichannel noise reduction algorithm on cochlear implant performance

Although cochlear implant (CI) users have enjoyed good speech recognition in quiet, they still have difficulties understanding speech in noise. We conducted three experiments to determine whether a directional microphone and an adaptive multichannel noise reduction algorithm could enhance CI performance in noise and whether Speech Transmission Index (STI) can be used to predict CI performance in various acoustic and signal processing conditions. In Experiment I, CI users listened to speech in noise processed by 4 hearing aid settings: omni-directional microphone, omni-directional microphone plus noise reduction, directional microphone, and directional microphone plus noise reduction. The directional microphone significantly improved speech recognition in noise. Both directional microphone and noise reduction algorithm improved overall preference. In Experiment II, normal hearing individuals listened to the recorded speech produced by 4- or 8-channel CI simulations. The 8-channel simulation yielded similar speech recognition results as in Experiment I, whereas the 4-channel simulation produced no significant difference among the 4 settings. In Experiment III, we examined the relationship between STIs and speech recognition. The results suggested that STI could predict actual and simulated CI speech intelligibility with acoustic degradation and the directional microphone, but not the noise reduction algorithm. Implications for intelligibility enhancement are discussed.