Speech enhancement using constrained spectral amplitude subtraction based on noncausal a priori SNR

Two gain forms of spectral amplitude subtraction are derived theoretically without neglecting the correlation of speech and noise spectrum during the period of a fralne. In the implementation, the constrained gain is expressed as a function of noncausal a priori SNR （Signal-to-Noise Ratio）. Noise and noncausal a priori SNR are estimated from the multitaper spectrum of the noisy signal with algorithms modified to be suitable for the multitaper spectruln. Objective evaluations show that in case of white Gaussian noise the proposed method outperforms some methods based on LSA （Log Spectral Amplitude） in terms of MBSD （Modified Bark Spectral Distortion）, segmental SNR and overall SNR, and informal listening tests show that speech reconstructed in this way has little speech distortion and musical noise is nearly inaudible even at low SNR.     

含噪语音信号频谱增强技术的统计方法研究与展望

基于单个麦克风的含噪语音信号频谱增强技术,一直受到有关工业和学术界的高度关注,其广泛应用于诸如语音识别、助听系统和免提终端通信等领域中。本文系统地讨论了含噪语音信号频谱增强系统设计的基本模块元素,并对诸如语音信号估计、语音信号出现概率估计、先验信噪比(SNR)估计和噪声功率谱估计等模块元素的统计技术与方法进行了较详细的讨论和描述。文中还讨论了含噪语音信号频谱增强算法的有关选择问题,并展望了其今后可能的研究与发展方向。     

Speech enhancement using super gauss mixture model of speech spectral amplitude

Most speech enhancement algorithms are derived by applying gauss hypothesis based on decor related speech samples. And recently it is accepted that the real pdf of speech spectral amplitude lies between the Laplace and Gamma amplitude approximation. This paper presents a supper gauss mixture model of speech spectral amplitude which can accurately approximates real speech spectral amplitude distribution. With this model, a new speech enhancement algorithm on the basis of MMSE estimator is derived. Simulations show that this algorithm has a better noise reduction performance and improve output SNR in the sense of speech segmentation.     

Combining Capon and APES for estimation of spectral lines

We propose combining the Capon and the APES spectral estimators for estimation of both the amplitude and the frequency of spectral lines. The so-obtained estimator does not suffer from Capon's biased amplitude estimates nor from APES' biased frequency estimates or resolution problem. Furthermore, the combined estimator is computationally simpler than APES and has about the same complexity as Capon. Numerical simulations are presented illustrating the increased performance.This work was supported in part by the Swedish Foundation for Strategic Research.     

A Bayesian estimation approach for speech enhancement using hiddenMarkov models

A Bayesian estimation approach for enhancing speech signals which have been degraded by statistically independent additive noise is motivated and developed. In particular, minimum mean square error (MMSE) and maximum a posteriori (MAP) signal estimators are developed using hidden Markov models (HMMs) for the clean signal and the noise process. It is shown that the MMSE estimator comprises a weighted sum of conditional mean estimators for the composite states of the noisy signal, where the weights equal the posterior probabilities of the composite states given the noisy signal. The estimation of several spectral functionals of the clean signal such as the sample spectrum and the complex exponential of the phase is also considered. A gain-adapted MAP estimator is developed using the expectation-maximization algorithm. The theoretical performance of the MMSE estimator is discussed, and convergence of the MAP estimator is proved. Both the MMSE and MAP estimators are tested in enhancing speech signals degraded by white Gaussian noise at input signal-to-noise ratios of from 5 to 20 dB     

Complex amplitude estimation in the known steering matrix and generalized waveform case

We consider a special growth-curve (SGC) model with a known steering matrix and generalized waveform in the presence of unknown interference and noise. Several estimators of the complex amplitude based on this model are derived, including the methods of approximate maximum likelihood (AML), minimum variance distortionless response (MVDR), and amplitude and phase estimation (APES). We analyze the statistical properties of these estimators and show that in the presence of temporally white but spatially correlated noise and interference, AML is asymptotically statistically efficient for a large snapshot number while MVDR and APES are asymptotically equivalent but not statistically efficient. Via several numerical examples, we also show that when the noise and interference are both spatially and temporally correlated, the APES estimator can achieve better estimation accuracy and exhibit greater robustness than the other methods.     

基于联合语音出现概率的先验信噪比估计算法

先验信噪比是语音增强的关键参数。该文分析了几种典型的先验信噪比估计算法,并得到这几种算法的统一形式,最后提出了基于联合语音出现概率的先验信噪比估计算法。测试结果表明,该算法在不引入音乐噪声的同时,平均段信噪比提高和平均对数谱距离等客观评价指标,都优于其它算法。     

一种LOFDM系统定时和频偏的盲估计算法

基于网格正交频分复用(LOFDM)信号的周期平稳性,该文提出一种LOFDM系统定时和载波频率偏差的盲估计算法。理论分析和仿真实验证实由该算法构造的估计器能够有效地对抗频率选择性慢时变信道引起的衰落;在信道噪声广义平稳的情况下,估计器性能与信噪比无关,于是估计器在低信噪比条件下也能很好地工作;另外,符号定时和频率偏差估计器的性能互不影响。     

Study on frequency estimation based on weighted least square method with three Fourier coefficients

In this paper, a sinusoidal signal frequency estimation algorithm is proposed by weighted least square method. Based on the idea of Provencher, three biggest Fourier coefficients in the maximum periodogram are considered, the Fourier coefficients can be written as three equations about the amplitude, phase, and frequency, and the frequency is estimated by solving equations. Because of the error of measurement, weighted least square method is used to solve the frequency equation and get the signal frequency. It is shown that the proposed estimator can approach the Cramer-Rao Bound （CRB） with a low Signal-to-Noise Ratio （SNR） threshold and has a higher accuracy.     

双通道能量差后滤波语音增强算法统计分析和改进

双通道能量差后滤波语音增强算法在语音通信系统的噪声抑制技术中有较好的应用前景,然而其理论性能和局限性还未得到充分研究。为此,本文采用统计分析方法研究了双通道能量差后滤波语音增强算法的性能,分析了相干性、平滑因子及噪声估计误差对算法的影响。理论和仿真结果表明,噪声估计误差和平滑因子严重影响该算法的降噪性能。依据此分析结果,本文提出一种基于非平稳噪声估计和功率谱自适应平滑的双通道能量差后滤波算法。测试结果表明,本文提出的算法在不增加语音失真的前提下,能更有效地抑制非平稳噪声,段信噪比提高(SegSNRI)和语音质量感知评估(PESQ)等客观评价指标都表明本文的算法优于其它几种经典的后滤波算法。      

Subspace-based noise variance and SNR estimation for MIMO OFDM systems

This paper proposes a subspace-based noise variance and Signal-to-Noise Ratio （SNR） estimation algorithm for Multi-Input Multi-Output （MIMO） wireless Orthogonal Frequency Division Multiplexing （OFDM） systems. The special training sequences with the property of orthogonality and phase shift orthogonality are used in pilot tones to obtain the estimated channel correlation matrix. Partitioning the observation space into a delay subspace and a noise subspace, we achieve the measurement of noise variance and SNR. Simulation results show that the proposed estimator can obtain accurate and real-time measurements of the noise variance and SNR for various multipath fading channels, demonstrating its strong robustness against different channels.     

一种基于语音和噪声统计信息改进相位估计的语音增强算法

为改善低信噪比环境下语音的质量,论文提出了一种改进相位估计的语音增强算法。算法首先根据语音和噪声频谱的统计模型的对称性得到用先验信噪比倒数形式表示的噪声频谱估计值,然后通过分析低信噪比条件下（0dB）相位估计对于幅度估计的重要性,利用噪声频谱估计值估计每一个频点的相位修正值,并给出了一种优化的先验信噪比估计算法,得到一种新的语音增强算法。由仿真实验给出的客观测试和非正式听音测试表明:该算法处理后取得了较好的效果,在抑制低信噪比语音增强所产生的音乐噪声的前提下,相比未改进相位估计的算法处理后的信号,语音失真度更小,语音质量有明显提高。      

一种基于短时谱估计和人耳掩蔽效应的语音增强算法

该文结合短时谱估计算法和人耳掩蔽效应提出了一种单通道语音增强算法。该算法在MMSE准则下采用了非固定参数的语音跟踪,并且引入人耳掩蔽效应动态的确定增强滤波器的传递函数以适应语音信号的变化。实验结果表明:该算法使降噪后的语音信号有较小的语音失真并且很好地抑制了音乐噪声。     

A spectral matching approach for parameter and spectral estimationof nonstationary rational processes

The parameter and spectral estimation problems of nonstationary signals are considered. The nonstationary signals are modeled as rational processes with time-varying parameters. The spectral matching approach, which was introduced by Friedlander and Porat (1984), is generalized to the nonstationary case and two new estimators, namely, the time-varying spectral matching estimator (TVSME) and the time-frequency spectral matching estimator (TFSME) are proposed. The proposed methods estimate the parameters of the time-varying rational model by fitting the parametric spectrum expression to an estimated time-frequency distribution of the signal. An approximate statistical analysis is given for both methods along with computer simulation results, illustrating the performance of the proposed estimators     

一种改进型MMSE语音增强方法

本文提出了一种改进型语音短时谱最小均方误差(MMSE)估计的增强方法。通过在每一帧及帧内每一频点对无音的概率(SAP)进行估计,得到Ephraim和MalahMMSE估计算法的改进形式。对增强后的语音客观和主观测试表明:在低信噪比条件下,相对于传统的谱减法和MMSE估计方法,这种改进的方法能更好的抑制背景噪声和残留的“音乐噪声”。     

基于传声器阵列语音增强方法研究

传声器阵列通过对拾取的多路语音信号进行分析与处理,能取得改进语音质量、消除背景噪声和提高语音可懂度等明显效果,现已成为语音信号增强的一个重要的研究领域。介绍了基于传声器阵列的自适应波束形成方法,该方法采用GSC结构基于TF-GSC的最优后置滤波算法。仿真实验结果表明,该自适应波束形成器对干扰有很好的消除作用,对阵元的增益误差、位置误差不敏感,可以取得较好的语音增强效果。     

采用复高斯分布模型的两步噪声幅度谱估计算法

噪声幅度谱估计是有效抑制外界噪声干扰、提高语音增强算法整体输出性能的重要环节。但目前针对该问题的研究相对较少,常用的语音激活检测算法只能在语音不存在阶段对噪声信号的幅度谱进行更新或估计,无法适用于更为复杂的非平稳噪声环境。为克服这一问题,本文基于噪声频谱的复高斯分布模型假设,提出了新型的两步噪声幅度谱估计算法。算法首先采用软判决技术计算噪声信号的功率谱,然后再结合复高斯分布条件下信号幅度谱和功率谱之间的数学关系间接地获取噪声幅度谱的估计。文中基于这一结论给出了两种估计算法,并在多种噪声环境下对它们的性能进行了仿真评估,其测试结果有效表明了提出算法优良的估计性能。      

基于DCT与维纳滤波的单通道语音增强算法

针对复杂噪声背景下的语音增强问题,基于离散余弦变换(DCT)和维纳滤波提出了一种新的单通道语音增强算法。该算法不依赖任何语音信号模型且无需对噪声的统计特性进行先验假定,它利用DCT域中连续时刻语音信号分量间的相关特性结合最小均方误差算法实现纯净语音分量的最优估计,弥补了一般算法仅依赖单帧带噪语音对语音分量估计得不足。多种噪声背景下的仿真结果表明,该算法在主观和客观测试中都具有良好的语音增强效果。     

Reduction of musical residual noise using perceptual tools with classic speech denoising techniques

Single channel enhancement techniques based on short-time spectral amplitude (STSA) estimation have the major drawback of generating an artificial and annoying residual noise with musical character, due mainly to the unwanted peaks in the denoised signal spectrum. The detection and reduction of spectral peaks which have a musical characteristic are the main objectives of this paper. The proposed perceptual technique to reduce musical residual noise operates as a post-processing. Based on human auditory properties, the perceptual post-processing is established in a number of steps. First, we detect musical peaks by comparing tonality coefficients in each critical band of both denoised signal and reference signal. Detected musical peaks are audible only if they exceed the clean speech masking threshold (MT). However, the clean MT is not available. It is estimated by modifying the Johnston model. Secondly, we reduce the musical residual noise by removing only audible musical peaks which exceed the estimated MT. The proposed method is tested and compared with classic STSA technique and perceptual techniques at various levels of white and colored noise. Results show the validity of the proposed technique.     

一种融合相位估计的深度卷积神经网络语音增强方法

在时频域的语音增强中,幅度估计和相位估计都是影响语音增强性能的重要因素.为了在基于深度学习的语音增强方法中融合对相位的估计,本文将含噪语音短时傅里叶变换（STFT）的实部和虚部特征作为两个通道输入深度卷积神经网络,通过建立一个同步估计纯净语音STFT的实部和虚部特征的多任务学习模型,实现了对幅度和相位的同步估计.实验结果表明,相比仅考虑幅度估计的方法,本文方法具有更好的噪声抑制能力,在低信噪比条件下,显著提高了语音增强性能.