相位均衡对耳机重放影响研究*

均衡耳机到鼓膜的传递函数可有效提高耳机声重放效果。通过主客观实验对比研究幅度均衡方法、相位均衡方法和幅度相位同时均衡方法,结果表明相比于幅度均衡,相位均衡对耳机瞬态响应的影响更大：当系统相位为线性时,瞬态响应衰减越快,线性相位对耳机重放提升音色有显著作用;而以平直幅频响应为目标的幅度均衡对提升耳机重放效果有限。     

相位均衡对耳机重放的影响

均衡耳机到鼓膜的传递函数可有效提高耳机声重放效果。通过主客观实验对比研究幅度均衡方法、相位均衡方法和幅度相位同时均衡方法,结果表明相比于幅度均衡,相位均衡对耳机瞬态响应的影响更大:当系统相位为线性时,瞬态响应衰减越快,线性相位对耳机重放提升音色有显著作用;而以平直幅频响应为目标的幅度均衡对提升耳机重放效果有限。     

最小方差无失真响应感知倒谱系数在说话人识别中的应用

研究最小方差无失真响应感知倒谱系数在说话人识别中的应用。提取最小方差无失真响应感知倒谱系数,对其进行高斯混合模型建模并采用联合因子分析的方法来拟合高斯混合模型中的说话人和信道差异,在美国国家标准技术研究院2008年说话人识别评测核心测试集上分别对最小方差无失真响应感知倒谱系数和传统的Mel频率倒谱系数进行测试。结果显示,两种不同特征的系统性能相当,采用线性融合方法后,在不同测试集上的等错误率相对下降了7.6%~30.5%,最小检测错误代价相对下降了3.2%~21.2%。实验表明,最小方差无失真响应感知倒谱系数能有效应用于说话人识别中,且与传统的Mel频率倒谱系数存在一定程度的互补性。      

基于仿生特征模式匹配的海洋动物声信号识别*

针对小数据量的海洋动物声信号混合识别,将声信号同态分析过程中的线性频率转换为Mel频率,得到模拟人耳听觉特性的Mel频率倒谱系数作为声信号的特征。按照声信号所属的物种建立特征模板,使用动态时间规整算法对待识别特征进行分类识别,并对特征库和识别算法进行优化。分别提取了6种鱼类、3种虾类、12种鲸类的Mel频率倒谱系数,为每个物种建立特征模板。分3次对3种、5种、6种鱼类进行识别,分别获得了100%、96.25%、94.68%的识别率。对6种鱼类、3种虾类、12种鲸类共21个物种进行混合识别,总识别率由87.56%提升至优化后的88.96%。实验结果表明,基于Mel频率倒谱系数和动态时间规整算法的海洋动物声信号混合识别能够在小数据量时获得较高的识别率,优化后的特征库和识别算法能够提升识别率。     

韵律特征在概率线性判别分析说话人确认中的应用

研究韵律特征在说话人确认中的应用。将整个韵律轨迹以固定段长和段移进行片段划分,并对其进行勒让德多项式拟合从而获取连续性的韵律特征,将特征映射到总变化因子空间,并用概率线性判别分析来补偿说话人和场景的差异。在美国国家标准技术研究院2010年说话人识别评测扩展核心测试集5的基础上加入噪声构造测试集,并分别对韵律特征和传统Mel频率倒谱系数进行测试。结果显示,随着信噪比的逐渐减小,Mel频率倒谱系数性能出现大幅度下降,而韵律特征性能相对比较稳定,两种特征融合后能使系统性能得到进一步提升,等错率和最小检测错误代价相对于Mel频率倒谱系数单系统最多能分别下降9%和11%。实验表明,韵律特征应用于说话人识别中具有较强的噪声鲁棒性,且与传统的Mel频率倒谱系数存在较强的互补性。      

韵律特征在概率线性判别分析说话人确认中的应用

研究韵律特征在说话人确认中的应用。将整个韵律轨迹以固定段长和段移进行片段划分,并对其进行勒让德多项式拟合从而获取连续性的韵律特征,将特征映射到总变化因子空间,并用概率线性判别分析来补偿说话人和场景的差异。在美国国家标准技术研究院2010年说话人识别评测扩展核心测试集5的基础上加入噪声构造测试集,并分别对韵律特征和传统Mel频率倒谱系数进行测试。结果显示,随着信噪比的逐渐减小,Mel频率倒谱系数性能出现大幅度下降,而韵律特征性能相对比较稳定,两种特征融合后能使系统性能得到进一步提升,等错率和最小检测错误代价相对于Mel频率倒谱系数单系统最多能分别下降9%和11%。实验表明,韵律特征应用于说话人识别中具有较强的噪声鲁棒性,且与传统的Mel频率倒谱系数存在较强的互补性。     

基于补偿距离估计的光纤振动信号特征提取算法

为进一步提高光纤振动信号模式识别的实时性及准确性,提出一种基于补偿距离估计技术的信号特征提取算法.算法借鉴人类听觉感知机理,对光纤传感振动信号提取Mel频率倒谱系数特征向量,采用补偿距离估计技术制定特征筛选策略实现特征评估与优化.实验结果表明,基于补偿距离估计技术的振动信号特征提取算法可以有效减少光纤传感系统中影响分类...     

基于动态单边自相关序列和频率规整线性预测的抗噪声语音识别

提出了一种既符合人耳听觉特性又具有良好抗噪性的语音特征分析方法。首先将单边自相关函数序列进行时间方向的平滑处理,提高单边自相关函数的抗噪性,然后用平滑后的单边自相关函数序列代替原信号进行频率规整的LPC分析,最后经倒谱变换得到该特征参数。数字语音识别实验证明:利用该特征参数的语音识别系统的识别性能优于MEL倒谱系数、LPC倒谱系数等传统的语音特征参数。     

改进的梅尔倒谱系数在低空飞行器特征提取中的应用*

梅尔倒谱系数特征提取技术依据人耳的感知特性将声信号从线性频域转换到梅尔域,在语音识别中得到广泛应用。该文将梅尔倒谱系数技术用于小型低空飞行器的声信号特征提取中,并针对螺旋桨驱动类的小型低空飞行器具有稳定的强谐波特性,对梅尔倒谱系数特征提取中使用的梅尔滤波器进行改进,通过对此类谐波处的线性频谱与梅尔谱转换曲线的斜率进行投影替换,提高滤波器对该谐波处信号的感知敏感度。仿真结果表明,使用改进的梅尔倒谱系数特征提取方法对小型低空飞行器进行特征提取时,能够得到更低的等误识率,并且在低信噪比环境中,改进的梅尔倒谱系数特征提取方法具有更好的抗噪能力。     

基于听觉感知特性的双耳音频处理技术*

自20世纪30年代引入立体声以来,人类对逼真的听觉体验一直进行着孜孜不倦的追求。双耳音频处理技术基于人耳听觉感知特性,利用计算机和数字信号处理等技术在听者双耳鼓膜处模拟出与真实场景相同的声压,以期给人以“身临其境”的体验,一直是音频信号处理领域的重要研究内容,特别是近两年随着虚拟现实等应用的蓬勃发展,得到更多关注。该文主要围绕双耳音频处理技术中所涉及的关键环节:双耳录音、双耳合成、耳机重放、扬声器重放、头跟踪等领域,以及相关典型应用场景进行较为系统的介绍,最后给出总结与展望。     

On the externalization of virtual sound images in headphone reproduction: a Wiener filter approach

In-head localization of sound images is a critical problem in headphone reproduction. The paper investigates the degree of externalization in terms of the distance of auditory images for various synthesis and reproduction cases. An effective binaural headphone system was constructed by way of binaural synthesis using head-related impulse responses and individual headphone equalization using Wiener filter theory. The headphone system designed had an average reproduction performance error of 2.4% for five subjects with a random noise input, and was used to perform some subjective tests with a set of virtual sources equally spaced and distanced from the center of each subject's head in the horizontal plane. The effects of individual and nonindividual binaural syntheses and those of equalized and nonequalized reproductions were separately investigated. In the tests, each subject was instructed to indicate the distance of auditory images. The results obtained demonstrate that individual equalization is important for externalization, and individual synthesis is important for consistent distance perception. Thus, a combined use of both individual equalization and individual synthesis resulted in externalized sound images of a consistent distance.     

利用经验模态分解的高频水声信号滤波方法*

研究了一种高频水声信号的滤波问题,提出了一种改进的经验模态分解加小波阈值滤波方法。首先对信号进行带通滤波处理,再进行经验模态分解,将分解得到的各个模态转换为频域信号,采用小波软阈值方法对这些频域信号进行滤波,最后对信号进行重构,并转换为时域信号。经数值仿真与试验数据验证表明此方法是可行有效的,与原基于经验模态分解的小波阈值滤波方法相比,本方法滤波效果较好:对不同输入信噪比的仿真信号进行滤波后,本方法的输出信噪比最大提高17.41 d B,滤波后所得信号与加噪前纯信号的相关系数最大提高0.90;对实验数据进行滤波后,不同时间段信号的相关系数最大提高0.62。     

Two-stage non-local means filtering with adaptive smoothing parameter

Non-local means (NLM) filtering is an efficacious algorithm in image denoising which searches the similar neighborhoods and estimates the pixel by averaging these neighborhoods. Some internal parameters such as patch size, search window size and smoothing strength have serious effects on filtering performance. This paper proposes an improved version of NLM by using weak textured patches based single image noise estimation and two-stage NLM with adaptive smoothing parameter. Our proposed method firstly applies weak textured patches based noise estimation to achieve the noise level of input noisy image. Then relying on the estimated noise level, we apply the first stage NLM with adaptive smoothing parameter to attain a basic denoised image. After that, the basic denoised image is refined by the second stage of NLM with smaller smoothing strength. Our experimental results show that the proposed algorithm outperforms the NLM and some NLM recent variants both in visual quality and numerical measures. Additionally, the potential halo effect is almost eliminated in the result images produced by our proposed method.     

The influence of time delay on forward masking when spatially separated sound image reproduced with headphone

With the virtual reproduction of headphone, the influence of delay time between masker and masked signal on forward masking when sound images are virtually spatial separated in azimuth is studied. The result shows that the thresholds tend to decrease dramatically with the increase in delay time. Furthermore, in case of spatial separation, there are differences of masking thresholds with the maximum about 15 dB. However, when the masker level is 70 dB, the difference of the thresholds increases a little as the delay time increases from 20 ms to 40 ms, which can be explained by the time integration effect and basilar membrane response.     

时间反转滤波器组多载波水声通信方法

针对水声滤波器组多载波(Filter Bank MultiCarrier,FBMC)传统均衡方法效率低、复杂度高等问题,提出了单阵元无源时间反转镜(Passive Time Reversal Mirror,PTRM)和虚拟时间反转镜(Virtual Time Reversal Mirror,VTRM)的均衡算法,该算法利用时间聚焦原理可以降低水声信道带来的符号间干扰(Inter Symbol Interference,ISI),并且不需要在FBMC符号中插入导频,进一步提高了通信速率.通过仿真和外场实验验证了水声FBMC系统采用单阵元PTRM和VTRM的有效性,结果表明,这两种均衡方法能够有效减少ISI,提高系统可靠性能。      

DWDM VSB modulation-format optical transmission: Effects of optical filtering and electrical equalization

The transmission of 40 Gb/s wavelength multiplexed channels under vestigial single side band modulation format is transmitted over long haul optically amplified fiber systems. Bit-error-rate (BER) of 10⁻¹² or better can be achieved across all channels. Optical filters are designed with asymmetric roll-off bands. Simulations of the transmission performance, BER versus receiver sensitivity are demonstrated with wavelength channel spacing of 20–40 GHz. An optical filter, whose passband is 28 GHz and 20 dB cut-off band, performs best for 40 Gb/s bit rate due to optimum filtering and minimum noise contribution. Furthermore the single-sideband property of VSB format can assist linear equalization by electronic processing. The transmission performance is accurately evaluated based on the eye opening using a fast statistical method based on an equivalent Gaussian probability density distribution (pdf) which is derived from multiple peaks pdf of distorted eye diagram.     

Smoothing effect in the broadband laser through a dispersive wedge

The propagation expression of a broadband laser passing through a dispersive wedge is derived on the basis of the Huygens-Fresnel diffraction integral. Smoothing effects caused by the phase perturbation of the dispersive wedge on the intensity profiles are investigated in detail. The phase perturbation of the dispersive wedge induces a relative transverse position shift between the diffraction patterns of different frequency components. The relative transverse position shift is of great benefit to the fill of the intensity peaks of some patterns in the valleys of others when these patterns are overlapped and thus the smoothing effect is achieved.     

马达加斯加玛瑙的光谱学特征研究

玛瑙在世界各地分布广泛,有关玛瑙矿物学和光谱学特征的研究颇丰,但对产自马达加斯加安楚希希市和马哈赞加省的玛瑙的光谱学和产地特征研究甚少。采用正交偏光显微镜、红外光谱和拉曼光谱仪对5块来自安楚希希市和马哈赞加省具有纹带结构的玛瑙样品进行光谱学特征分析。结果表明：这两个产地玛瑙的主要矿物成分一致,均为α-石英和斜硅石。来自马哈赞加省的样品靠近中心的白色、红色部分斜硅石的特征拉曼峰（501cm-1）缺失,根据前人研究发现是与玛瑙内部的结构水之间发生脱水反应形成中性水分子,而这些中性水分子与斜硅石反应生成了石英颗粒有关。然而两产地玛瑙的致色矿物不同,马哈赞加省玛瑙棕色和红色部分具有222, 294, 410以及1 316 cm-1的特征拉曼位移峰,与赤铁矿的拉曼特征峰相符,确定这部分的致色矿物是赤铁矿,而安楚希希市玛瑙黄色部分只具有400和1 160 cm-1的特征峰,与六方纤铁矿的拉曼特征峰相符,因此此部分的致色矿物是六方纤铁矿（δ-FeOOH）。部分来自安楚希希市的玛瑙具有少量的杂质矿物,通过拉曼测试发现杂质矿物的成分也是六方纤铁矿（δ-FeOOH）和α-石英。结合六方纤铁矿的特性推测来自安楚希希市的四块样品是处于低温（<80 ℃）、Fe2+含量很少（β<0.03）且空气不流通的偏酸性的环境当中形成的。红外光谱测试发现两个产地的玛瑙均有液态水（3 400 cm-1左右）和结构水（3 750～3 500 cm-1）的红外特征峰,安楚希希市的结构水红外特征峰位于3 740 cm-1处,可推测该结构水位于结构缺陷处。结合马达加斯加玛瑙的光谱学特征和纹带结构的讨论,对推测马达加斯加玛瑙的形成机制有着重要的意义;并且六方纤铁矿（δ-FeOOH）可以作为区别于其他产地的标型矿物,具有产地特征意义;同时为研究马达加斯加玛瑙形成的地质环境提供了数据支持。