基于多频带谱减法的抗噪声语音识别研究

为了减少在噪声环境下测试条件与训练条件不匹配导致的语音识别性能下降,提出了一种结合多频带谱减法的抗噪声语音识别系统。首先提取带噪语音的前几帧作为估计的噪声信号,将带噪语音、估计的噪声信号按频率划分M个互不相交的频带,然后根据每个频带内带噪语音与估计的噪声信号的性噪比,来确定该频带噪声的谱减参数。语音增强作为前端处理,与语音识别器级连构成抗噪声语音识别系统。通过实验仿真表明,基于多频带谱减法的抗噪声语音识别系统在不同信噪比不同类型的噪声下,识别性能明显优于基本谱减法。     

噪声环境下语音能量的MMSE估计及其在语音识别中的应用

语音识别中,语音信号能量是一个重要的识别特征参数.在噪声环境下,语音信号能量产生的畸变是导致识别率下降的原因之一.本文依据RASTA-PLP分析的增益参数和最小均方误差估计准则,给出了带噪语音能量和干净语音信号能量的线性映射关系.在中文连续语音识别中的应用表明,利用该映射关系得到的干净语音能量估计,可使信噪比为5dB的白噪声环境下的识别正确率提高21.36%.     

分数傅里叶变换域上含噪语音的联合滤波

噪声是影响语音识别和说话人识别性能的主要因素,目前常用的降噪方法多是针对平稳噪声的,而针对非平稳噪声的降噪方法很少。而在实际环境中,通常的噪声是非平稳的。本文将含噪语音变换到分数傅立叶域上,提出了一种在分数傅立叶变换域上进行线性最优滤波和中值滤波的联合滤波降噪方法。实验结果表明,该方法对含非平稳噪声的语音的降噪效果明显优于维纳滤波,能够有效地降低非平稳噪声的影响,提高非平稳噪声环境下的语音识别和说话人识别性能。     

非平稳噪声环境下的噪声估计算法

通过对噪音和语音频谱的分析,针对航空背景噪声的特性,提出一种用于语音增强的新的噪声估计算法。通常的噪声估计一般利用语音端点检测方法,取噪声段的谱平均值作为待估计的噪声谱,但该方法在信噪比较低时性能下降严重。笔者提出的基于频率段能量比的噪音谱估计方法,不依赖于语音端点检测而直接由语音帧来估计噪音谱,通过计算一帧语音中各频率段中能量比,以判断该帧是否含有语音来修正噪声谱估计的计算因子。算法提高了谱减法的适用范围,还在一般谱相减方法的基础上提出了改进的谱相减算法。     

基于LPC美尔倒谱特征的带噪语音端点检测

复杂的噪声环境是语音识别系统在实际应用中性能下降的原因之一,识别预处理中的带噪端点检测作为关键技术,其性能的优劣某种程度上决定了识别率的高低。笔者提出了基于LPC美尔倒谱特征的带噪端点检测方法,对语音信号分高低频段分别提取IPC美尔倒谱特征分析,根据Mel倒谱距离判决,采用自适应噪声估计,实验结果表明,该方法计算效率较高,低信噪比下有较好的检测性能。     

一种基于概率模型和倒谱差分的特征补偿算法

本文详细给出了概率模型中引入倒谱预测值的动态相关性来进行特征补偿的方法.该方法采用期望最大化(EM)算法来估计联合分布参数,基于语音和噪声的先验概率密度、在倒谱域对语音特征参数进行最小均方误差预测(MMSE),以提高语音识别精度.不同噪声环境和不同信噪比下的实验结果表明,本文方法能有效提高噪声环境下的中文连续语音识别的正确率.     

基于子带加权和GMM的鲁棒性话者确认的研究

噪声鲁棒性是话者确认系统实用化的关键问题之一,本文设计了一种基于子带加权和GMM的话者确认系统,该系统将语音谱分为若干子带,采用基于短时能量分布的算法估计各子带噪声强度,并根据噪声强度来进行子带加权,最终生成具有更高鲁棒性的语音特征,语音识别模型采用简化的GMM.实验表明,上述方法能有效提高话者确认系统的性能,增强其噪声鲁棒性,而且在低噪声环境下,仍能保证系统性能不下降.     

室内声环境中的语音端点检测自适应算法

室内场所是语音识别技术的一种典型应用环境,传统的端点检测研究多考虑噪声的影响,忽略室内混响的影响,研究证明室内混响对端点检测和识别效果能造成显著的负面影响.通过研究短时能量和短时自相关序列( RAS),提出了一种自适应的端点检测方法.可以通过估计噪声段短时能量来适应平稳噪声干扰环境,并能修正含混响语音的检测终点.端点检测和语音识别实验结果表明,本方法在平稳噪声和室内混响声环境下具有良好的性能.     

基于改进语音特征提取方法的语音识别

在分析语音特征提取方法基础上提出一种改进组合算法,并采用HMM声学模型和Viterbi算法进行模式训练和识别.实验结果表明,该算法在噪声环境中具有较好的鲁棒性,能有效提高噪声环境下中文连续语音识别的正确率,增强语音识别整体性能,因此在噪声环境下的语音识别系统中具有一定的实用价值.     

噪声自适应的多数据流复合子带语音识别方法

首先针对现有丢失数据语音识别技术中的边缘化(marginalisation)技术在特征运用上的局限,提出了一种倒谱特征分量的可靠性估计方法,将边缘化技术推广到常用的倒谱语音识别系统中; 然后利用基于全带和子带倒谱特征的边缘化识别器在不同噪声中的互补性能,提出了一种噪声自适应的多数据流复合子带语音识别方法。实验结果表明,所提识别方法可以自适应地选出全带和子带数据流中受噪声影响较小者并以之为主要依据进行识别,有效地提高了识别系统在多变噪声环境中的鲁棒性。     

System parameter estimation with input/output noisy data andmissing measurements

An investigation is undertaken to examine the parameter estimation problem of linear systems when some of the measurements are unavailable (i.e., missing data) and the probability of occurrence of missing data is unknown a priori. The system input and output data are also assumed to be corrupted by measurement noise, and the knowledge of the noise distribution is unknown. Under the unknown noise distribution and missing measurements, a consistent parameter estimation algorithm [which is based on an l_p norm iterative estimation algorithm-iteratively reweighted least squares (IRLS)] is proposed to estimate the system parameters. We show that if the probability of missing measurement is less than one half, the parameter estimates via the proposed estimation algorithm will converge to the true parameters as the number of data tends to infinity. Finally, several simulation results are presented to illustrate the performance of the proposed l _p norm iterative estimation algorithm. Simulation results indicate that under input/output missing data and noise environment, the proposed parameter estimation algorithm is an efficient approach toward the system parameter estimation problem     

Direction-of-arrival estimation based on spatial–temporal statistics without knowing the source number

Direction-of-arrival (DOA) estimation is a central problem in array processing and has a variety of applications. In this paper, a new algorithm for finding DOAs of multiple temporally correlated signals is devised. The proposed approach is based on the joint diagonalization structure of a set of spatio-temporal correlation matrices. Unlike the subspace-based DOA estimators, it is not necessary to estimate the noise or signal subspace explicitly. Moreover, the proposed method can provide the spatial spectrum and estimate the DOAs even when the number of sources is not known a priori. Interestingly, it is revealed that the well-known MUSIC method is a special case of our algorithm. Simulation results validate that the developed approach is superior to conventional DOA estimators in terms of resolution capability, estimation accuracy, and robustness against array model errors.     

A Neural Network Approach to Blind Estimation of PN Spreading Sequence in DS/SS Signals

As the power spectrum density of direct sequence spread spectrum (DS/SS or DS), signals is often much lower than the noises’, the DS signals have the ability to resist interception and interference. But in point of fact, if we have get some parameters of the DS signals, including information symbol period, chip period of the pseudo noise (PN) sequence, we can estimate the PN sequence in DS signals blindly, this has great value for management or interception of DS communications. Ref.[1…     

一种Alpha稳定分布噪声下的体积目标长度估计方法

为了解决Alpha稳定分布噪声环境下运动舰船目标的长度估计问题,该文借鉴非线性变换抑制脉冲噪声以及多普勒目标运动特性估计思想,提出基于广义时频分析(G-TFA)和最小二乘估计的运动目标长度估计方法。该方法首先利用G-TFA获取Alpha稳定分布噪声环境下运动目标的多普勒频率,然后利用最小二乘方法估计出目标航速和不同位置的横正时刻,最后利用上述估计结果计算目标长度。以广义Winger-Ville分布(G-WVD)为例,从理论上推导了G-TFA在Alpha稳定分布噪声环境下具有提取目标多普勒特征的能力,并通过仿真实验验证了该算法在中低混合信噪比下的稳健性。与现有算法相比,该文所提算法不需要估计噪声特征指数,算法性能优于基于传统时频分析的估计方法。     

乘性和加性噪声相关背景下的二维谐波频率估计

本文利用二维循环统计量方法对乘性噪声之间相关、乘性噪声和加性噪声之间也相关这种复杂噪声背景下的谐波恢复问题进行了研究.首先,提出了二维噪声互可混的概念,用它来体现多个二维噪声之间的关系;然后,在乘性噪声为非零均值时,定义了二维循环均值来估计信号频率.在乘性噪声和加性噪声为零均值时,定义特殊的二维六阶时间平均多矩谱切片来估计信号频率.仿真实验证明了算法的有效性.     

Estimation and detection in non-Gaussian noise using higher orderstatistics

One of the primary applications of higher order statistics has been for detection and estimation of nonGaussian signals in Gaussian noise of unknown covariance. This is motivated by the fact that higher order cumulants of Gaussian processes vanish. We study the opposite problem, namely, detection and estimation in nonGaussian noise. We estimate cumulants of nonGaussian processes in the presence of unknown deterministic and/or Gaussian signals, which allows either parametric or nonparametric estimation of the covariance of the nonGaussian noise. Our approach is to augment existing second-order detection methods using cumulants. We propose solutions for detection of deterministic signals based on matched filters and the generalized likelihood ratio test which incorporate cumulants, where the resulting solutions are valid under either detection hypotheses. This allows for single record detection and obviates the need for noise-only training records. The problem of estimating signal strength in the presence of nonGaussian noise of unknown covariance is also considered, and a cumulant-based solution is proposed which uses a single data record. Examples are used throughout to illustrate our proposed methods     

克服噪声不确定度的拟合优度检验频谱感知算法

针对已有的基于拟合优度(GoF)检验的频谱感知算法易受到噪声不确定度影响的问题,利用矩估计法或特征分解估计法对噪声方差进行实时估计,将采样数据处理为标准正态分布的信号,最后通过GoF检验来感知主用户的存在性.在减小GoF算法复杂度的同时,克服了噪声不确定度对算法性能的影响,仿真结果也表明了所提算法的有效性.     

Estimation of dynamically evolving ellipsoids with applications to medical imaging

The estimation of dynamically evolving ellipsoids from noisy lower-dimensional projections is examined. In particular, this work describes a model-based approach using geometric reconstruction and recursive estimation techniques to obtain a dynamic estimate of left-ventricular ejection fraction from a gated set of planar myocardial perfusion images. The proposed approach differs from current ejection fraction estimation techniques both in the imaging modality used and in the subsequent processing which yields a dynamic ejection fraction estimate. For this work, the left ventricle is modeled as a dynamically evolving three-dimensional (3-D) ellipsoid. The left-ventricular outline observed in the myocardial perfusion images is then modeled as a dynamic, two-dimensional (2-D) ellipsoid, obtained as the projection of the former 3-D ellipsoid. This data is processed in two ways: first, as a 3-D dynamic ellipsoid reconstruction problem; second, each view is considered as a 2-D dynamic ellipse estimation problem and then the 3-D ejection fraction is obtained by combining the effective 2-D ejection fractions of each view. The approximating ellipsoids are reconstructed using a Rauch-Tung-Striebel smoothing filter, which produces an ejection fraction estimate that is more robust to noise since it is based on the entire data set; in contrast, traditional ejection fraction estimates are based only on true frames of data. Further, numerical studies of the sensitivity of this approach to unknown dynamics and projection geometry are presented, providing a rational basis for specifying system parameters. This investigation includes estimation of ejection fraction from both simulated and real data.     

Cumulant-based method for bearing estimation in the presence ofnonGaussian noise

Bearing estimation algorithms based on the cumulants of array data have been developed to suppress additive spatially correlated Gaussian noises. In practice, however, the noises encountered in signal processing environments are often non-Gaussian, and the applications of those cumulant-based algorithms designed for Gaussian noise to non-Gaussian environments may severely degrade the estimation performance. The authors propose a new cumulant-based method to solve this problem. This approach is based on the fourth-order cumulants of the array data transformed by DFT, and relies on the statistical central limit theorem to show that the fourth-order cumulants of the additive non-Gaussian noises approach zero in each DFT cell. Simulation results are presented to demonstrate that the proposed method can effectively estimate the bearings in both Gaussian and non-Gaussian noise environments     

Active bearing estimation from a mobile two-dimensional array of sensors

A new approach for active-sonar target detection and bearing estimation from a mobile two-dimensional array of sensors operating in a predominantly noisy environment is presented. Sensor-level adaptive noise cancellation featuring an unconventional method for reference-noise estimation is the key preprocessing step in the proposed approach. A signal-subspace algorithm resulting from two-stage optimisation based on a generalised eigendecomposition of the signal plus (residual) noise covariance matrix is employed to estimate the bearing of the detected target. Simulation results conclusively demonstrate that the proposed scheme is capable of performing target detection and the subsequent two-dimensional bearing estimation with a high degree of reliability at signal-to-noise power ratios as low as -70 and -40-dB, respectively.