A blind algorithm for reverberation-time estimation using subband decomposition of speech signals

An algorithm for blind estimation of reverberation time (RT) in speech signals is proposed. Analysis is restricted to the free-decaying regions of the signal, where the reverberation effect dominates, yielding a more accurate RT estimate at a reduced computational cost. A spectral decomposition is performed on the reverberant signal and partial RT estimates are determined in all signal subbands, providing more data to the statistical-analysis stage of the algorithm, which yields the final RT estimate. Algorithm performance is assessed using two distinct speech databases, achieving 91% and 97% correlation with the RTs measured by a standard nonblind method, indicating that the proposed method blindly estimates the RT in a reliable and consistent manner.     

结合卷积神经网络与混响时间注意力机制的混响抑制

提出一种结合卷积神经网络的编解码器模型和混响时间注意力机制的混响抑制算法,该算法通过编解码器模型实现混响抑制,并利用混响时间注意力机制克服混响环境变化对混响抑制效果的影响。该算法在编码器中使用具有不同大小的卷积核来处理混响语音幅度谱,从而获得包含多尺度上下文信息的编码特征;通过引入注意力模块,实现在不同的混响时间环境中选择性地使用不同权重的编码特征生成加权特征;最后,在解码器中使用加权特征来重建混响抑制后的语音信号幅度谱。在模拟和真实的混响环境下,该算法相对于基线系统在语音混响调制能量比上分别取得了0.36 dB和0.66 dB的提升。实验结果表明,该算法可以适应不同混响环境的变化,相对基线系统在真实混响环境下具有更高的鲁棒性。      

Sound propagation in long enclosures with a vertical or inclined branch

It is of interest to investigate the sound propagation in long enclosures with a vertical/inclined branch, since this may be relevant to long enclosures in practical applications, such as in high-speed railway tunnels. However, much research on sound propagation in long enclosures available is concerned with straight long enclosures. In this paper, the sound pressure level (SPL) attenuation, early decay time (EDT), and reverberation time (RT30) of long enclosures with vertical/inclined branches have been studied by comparing experimental results from physical scale models, of such enclosures with those for a straight long enclosure. It has been found that, compared with the straight long enclosure, the inclined branch makes more-difference to the SPL, EDT and RT30 of the long enclosure than the vertical branch, even though both kinds of branches have the equivalent volume.     

运动平台低频混响的方位-多普勒谱特性研究

运动平台混响场的空频耦合结构对于改善主动声呐低速目标检测性能是十分关键的。本文通过对波束域混响采用高分辨谱分析方法(如MUSIC法和Capon法)在一定程度上突破了发射脉宽的制约,可以更为精确地给出低频混响场的方位-多普勒扩展估计。对南海实验数据的分析表明浅海低频混响场精确地符合运动多普勒规律,并且在一定距离方位分辨单元内的具有混响强度越高,频率扩展越小的特点。进一步的,利用运动多普勒关系给出偏航条件下左右舷混响在方位多普勒平面上沿不同曲线分布的规律,实验数据与理论预测符合良好。      

Blind single-channel suppression of late reverberation based on perceptual reverberation modeling

A blind method for suppressing late reverberation from speech and audio signals is presented. The proposed technique operates both on the spectral and on the sub-band domains employing a single input channel. At first, a preliminary rough clean signal estimation is required and for this, any standard technique may be applied; however here the estimate is obtained through spectral subtraction. Then, an auditory masking model is employed in sub-bands to extract the reverberation masking index (RMI) which identifies signal regions with perceived alterations due to late reverberation. Utilizing a selective signal processing technique only these regions are suppressed through sub-band temporal envelope filtering based on analytical expressions. Objective and subjective measures indicate that the proposed method achieves significant late reverberation suppression for both speech and music signals over a wide range of reverberation time (RT) scenarios.     

利用正弦调频信号的宽带速度敏感特性抑制混响

有源声呐探测水中低速目标时,混响与目标回波的频率接近,目标回波与混响分离困难.根据运动目标与产生混响的散射体在速度域的差异,利用正弦调频信号(SFM)的速度敏感特性实现目标回波与混响的有效分离.推导了 SFM信号宽带混响抑制能力与速度分辨模糊之间的约束关系,建立了波形优化设计的搜索模型.通过多普勒滤波处理将数据从频率域...     

基于多重核学习支持向量回归的混沌时间序列预测

鉴于标准支持向量回归应用于混沌时间序列预测时经常会遇到诸如核函数及其参数难以确定的问题,提出了多重核支持向量回归的方法.通过在混合核空间求解二次约束下的二次规划问题,实现多重核支持向量回归算法.该算法不仅可以减少支持向量的个数,而且能够提高预测性能.最后将该方法运用到Lorenz, Henon和Mackey-Glass混沌时间序列预测,仿真结果表明该方法能够有效地提高预测精度,增强预测模型的泛化性能. 关键词： 混沌时间序列 支持向量机 多重核学习 优化     

非线性内波存在下的浅海远程混响模型

研究了浅海非线性内波对远程混响场的影响。通过分析非线性内波活动引起声源到海底散射元以及散射元到接收点之间的声传播变化,给出了非线性内波引起远程混响强度变化的表示,建立了非线性内波存在下的浅海远程混响模型,数值计算了非线性内波运动引起远程混响强度的变化。理论和数值计算表明,非线性内波的活动能够引起远程混响强度的变化,在某些情况下会导致远程混响强度增强。通过讨论非线性内波引声简正波的耦合效应,给出了其引起远程混响强度增强的原因。      

房间脉冲响应模拟法及混响时间盲估计应用

在构建混响语声数据集时,由于缺乏真实长混响房间脉冲响应且模拟的房间脉冲响应与真实不符,因而导致数据驱动的混响时间盲估计模型性能下降。提出了一种基于条件生成对抗网络的房间脉冲响应模拟法,该方法利用真实的房间脉冲响应训练条件生成对抗网络,可以根据指定的混响时间模拟更加真实的房间脉冲响应。使用不同方法模拟的房间脉冲响应构建训练集用于训练盲估计模型,通过声学实验评估模型性能。实验结果表明,由该方法模拟的房间脉冲响应训练的估计模型在不同信噪比下均具有最小的均方根误差且在长混响情况下显著优于其他模型。     

水下目标回波盲分离性能的瞬时频率特征评价方法

在水下主动声呐目标回波与混响盲分离中,针对分离结果顺序不确定性以及缺乏分离有效性衡量手段的问题,提出了以信号瞬时频率特征为指标的盲分离性能评价方法。推导了目标回波与混响的时频分布特性,理论表明目标回波在瞬时频率序列的中心偏离程度以及整体随机程度上低于混响,据此提取信号的瞬时频率方差与瞬时频率熵两种信号特征,并将二者作为从盲分离结果中识别目标回波的依据。海试数据结果表明,在盲分离得到的所有分离信号中,目标回波具有最小的瞬时频率特征值,并且该特征值越小,目标回波与混响的盲分离程度就越高。      

Blind estimation of reverberation time

The reverberation time (RT) is an important parameter for characterizing the quality of an auditory space. Sounds in reverberant environments are subject to coloration. This affects speech intelligibility and sound localization. Many state-of-the-art audio signal processing algorithms, for example in hearing-aids and telephony, are expected to have the ability to characterize the listening environment, and turn on an appropriate processing strategy accordingly. Thus, a method for characterization of room RT based on passively received microphone signals represents an important enabling technology. Current RT estimators, such as Schroeder's method, depend on a controlled sound source, and thus cannot produce an online, blind RT estimate. Here, a method for estimating RT without prior knowledge of sound sources or room geometry is presented. The diffusive tail of reverberation was modeled as an exponentially damped Gaussian white noise process. The time-constant of the decay, which provided a measure of the RT, was estimated using a maximum-likelihood procedure. The estimates were obtained continuously, and an order-statistics filter was used to extract the most likely RT from the accumulated estimates. The procedure was illustrated for connected speech. Results obtained for simulated and real room data are in good agreement with the real RT values.     

Investigation of Chinese word recognition scores of children in primary school classroom with different speech sound pressure levels

Chinese word recognition (CWR) test was conducted by grades 3 and 5 children under the different conditions of reverberation time (RT), background noise level (BNL) and speech sound pressure level (SSPL) in three primary-school classrooms. The CWR scores and signal to noise ratios (SNRs) have been obtained at listening positions. Results show that the CWR score for grades 3 and 5 children increases with increase of SSPL, decrease of RT or increase of age, but it decreases with increase of BNL under the same conditions. For a mixed noise of 56 dBA (speech-spectrum-like noise and ambient noise), the CWR scores in the classroom for grades 3 and 5 children reach a peak at SNR of 15–20 dBA under the same RT and age of children condition. For the natural ambient noise, the CWR score for grades 3 and 5 children gradually increases with increase of the SNR. The high SSPL could not guarantee good CWR for children in classroom, which also depends on RT and BNL in classroom. When the classroom has long RT or high BNL, the increase of SSPL would not be necessarily to achieve better CWR. The novelty of the present study is to further evaluate and confirm the results under environments of real classrooms (not simulated room in laboratory).     

Study on sound propagation in long enclosures with a vertical or inclined branch of different dimensions

As there may be some branches in long enclosures, such as high-speed railway and freeway tunnels, it will make a difference in the sound propagation in long enclosures if we give more consideration to different designs of the branches at the onset of their construction. However, most researches on the sound propagation in long enclosures available are concerned with straight long enclosures. In this paper, the sound pressure level (SPL) attenuation, early decay time (EDT), and reverberation time (RT30) of long enclosures with a vertical or inclined branch of different dimensions have been studied by comparing experimental results from physical scale models of such enclosures with those of the straight long enclosure. This experimental investigation gives interesting results on the behavior of sound propagation in long enclosures with a vertical or inclined branch of different dimensions. When conducting engineering design for long tunnels of the high-speed railway or freeway, it would be appropriate to consider designing the branch of the tunnels into a widened left inclined or right inclined one to provide a better relative SPL than that of other branches with different dimensions. This study further reveals that with an inclined branch of different dimensions, the EDT of the long enclosure will be different. At both 500 and 1000 Hz 1/3 octave bands, the EDT is the shortest when the long enclosure has a left inclined branch and an extended left inclined branch, while the EDT is the longest when the long enclosure has a widened vertical branch. Generally speaking, in the near field (i.e. the source-receiver distance is shorter than the threshold distance), the EDT measurements are similar, while in the far field (i.e. the source-receiver distance exceeds the threshold distance), the EDT is the longest for the straight long enclosure, second longest for the long enclosure with a widened inclined branch, third longest for the long enclosure with a vertical branch, and the shortest for the long enclosure with an extended inclined branch, which is similar to the trend with RT30.     

Statistical comparison of reverberation times measured by the integrated impulse response and interrupted noise methods, computationally simulated with ODEON software, and calculated by Sabine, Eyring and Arau-Puchades’ formulas

This paper evaluates several procedures to determine the reverberation time, RT, in a classroom. These procedures are: (1) measurement by the integrated impulse response method, (2) measurement by the interrupted noise method, (3) computer simulation using ODEON Version 9.0 software, and (4) calculations using the Sabine, Eyring, and Arau-Puchades formulas. The resulting data are analyzed statistically to verify their similarity. No statistical difference was found between the values obtained by the two measuring methods. The computer simulation produced accurate data. The best formula for calculating RT in the classroom in question is Eyring’s formula.     

Generalized fractional S-transform and its application to discriminate environmental background acoustic noise signals

We propose a modification of S-transform (ST) by changing the kernel of Fourier transform (FT) with that of fractional Fourier transform (FRFT) and call it generalized fractional ST (GFST). The FRFT is a generalization of FT and it has been shown more useful than the FT for signals with changing frequencies such as chirp signals. The proposed GFST is applied to analyze and classify different environmental background sound mixed with speech signal in the form of additive noise. The simulation results demonstrate that Euclidean distance between the feature vectors computed from generalized fractional ST corresponding to different background noise is increased as compared to ST for the same set of feature vectors and signals.     

Acoustical design of the Tokyo Opera City (TOC) concert hall, Japan

The Tokyo Opera City concert hall seats 1632, volume 15 300 m3, and reverberation time, with audience and orchestra, 1.95 s. As part of the design process, measurements on CAD computer and 1:10 wooden models of the hall and full-sized materials samples were conducted over a 5-yr. period. The hall in plan is rectangular. The ceiling is a distorted pyramid, with its peak 28 m above the main floor and nearer the stage than the rear of the hall. This unique shape was analyzed on the models so that all interior surfaces combine to distribute sources on the stage uniformly over the seating areas and to yield optimum values for reverberation time (RT), early decay time (EDT), interaural cross-correlation coefficient (IACCE3), bass ratio (BR), initial-time-delay gap (ITDG), strength (G), and sound diffusion index (SDI) [for definitions see L. Beranek, Concert and Opera Halls: How They Sound (Acoustical Society of America, Woodbury, NY, 1996)]. On the long ceiling facing the stage, Schroeder QRD diffusers provide diffusion, eliminate a possible echo, and strengthen lateral reflections. Performers and critics judge the acoustics excellent.     

水声信号稀疏重构高分辨角度-多普勒成像

针对傅氏空时二维谱估计分辨率低以及声呐空时采样数据样本数不足给角度-多普勒成像带来困难的问题,提出一种水声信号稀疏重构的高分辨角度-多普勒成像方法和抗混响空时滤波器的稀疏重构方法.该方法在声呐阵列单测量向量的极少观测样本条件下,建立阵列信号的空时稀疏表示模型,应用稀疏表示的匹配追踪算法和基追踪算法重构回波与混响的高分辨...     

混响室条件下环境场强测量位置选取准则

针对在混响室条件下进行辐射敏感度测试时电磁环境场强难以度量的问题,分析了混响室内电场强度的统计特性,理论推导了有损混响室内场强空间相关系数表达式,并通过仿真分析验证了表达式的正确性。根据表达式得出有损混响室内相对距离大于半波长的两点间场强相互独立的结论,并由此推断出,混响室内环境场强测试位置与受试设备(EUT)距离应大于0.5倍波长,且应以此作为测量位置选取准则。采用仿真计算与实验验证相结合的方法分析了相对误差随测试点与EUT间相对距离的变化规律,对该准则进行了验证,结果表明,当测试点与EUT相对距离大于0.5时,相对误差能够保证在1.5 dB以内,可应用于实际混响室环境场强测试。     

High-resolution angle-Doppler imaging by sparse recovery of underwater acoustic signals

The low resolution of Fourier two-dimensional spatial temporal spectrum estimation and the insufficient sample size of sonar space time sampling data often caused difficulties in high-resolution space time spectrum estimation. Aiming to solve this problem, we proposed a high-resolution angle-Doppler imaging method and designed an anti-reverberation space time filter based on the sparse recovery of underwater acoustic signals. The proposed imaging method established the spatial temporal sparse representation model of array signal under the condition of few observation samples of single measurement vector(SMV), and reconstructed the high-resolution angle-Doppler profile of echo and reverberation through the matching pursuit(MP) algorithm and basis pursuit(BP) algorithm. By utilizing the space time distribution characteristics of echo and reverberation and the prior information of sonar rangecell under test(RUT), a reverberation dictionary composed of special space time guidance vectors was designed, and was used to reconstruct the reverberation and formed an anti-reverberation space time filter to suppress the reverberation interference of angle-Doppler plane. Computer simulations indicated that, under two conditions of forward and side-view array of moving sonar, the single measurement vector of sonar array was successfully used in the reverberation background to reconstruct the high-resolution angle-Doppler profile of low-speed moving target multi-spot echo. Its frequency resolution had a better performance than the Fourier resolution and the angle resolution broke through the Rayleigh limit of the array, the resolution was obviously superior to the Fourier space time spectrum estimation.     

基于分数阶傅里叶变换混响抑制的目标回波检测方法

提出了一种声呐发射信号为线性调频信号时,基于分数阶傅里叶变换混响抑制的目标回波信号检测方法。通过计算混响的缓慢时变包络,对混响进行时间域的平稳化处理。对平稳化的混响信号滑动窗截取,对截取信号进行分数阶傅里叶变换,然后在分数阶傅里叶域进行滤波处理,再进行逆分数阶傅里叶变换恢复出时间域信号,最后输出该信号的能量。当滑动窗截取到目标回波信号时,窗内的混响噪声得到抑制,系统输出目标回波的能量,从而实现混响背景下的信号检测。通过计算机仿真和湖试实验结果,表明所提方法可以准确的在混响背景下检测到目标回波信号,并且在混响噪声背景条件下相比于匹配滤波器具有更好的检测性能。