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。      

基于噪声追踪的二值时频掩蔽到浮值掩蔽的泛化算法

虽然浮值掩蔽比二值掩蔽有更好的语音分离效果,但是由于理想浮值掩蔽难以直接估计,现有的语音分离系统通常以理想二值掩蔽估计作为计算目标。我们提出了一个二值掩蔽到浮值掩蔽的泛化算法。由于实现浮值掩蔽估计的关键在于噪声能量追踪,我们首先采用指数分布刻画以混合谱和噪声能量以混合能量及二值掩蔽为观测的条件分布。其次,采用高斯马尔柯夫条件随机场刻画噪声估计在连续几帧内的关联。最后,采用马尔柯夫链-蒙特卡洛计算噪声能量最小均方误差估计并进一步计算浮值掩蔽。实验表明,相比于基于二值掩蔽估计的常规算法,我们所提出的算法在信噪比增益和客观感知质量两方面都有显著提高。     

分段线性光谱分类器的优化算法设计

快速准确地识别污染气体种类是光谱法环境监测技术对分类器的基本要求。分段线性分类器简单、计算量小,可以较好的逼近非线性分界面。文章根据最大化分类间隔的思想,结合分段线性分类器和线性支持向量机,设计了单边分段线性分类器优化算法。对某气体模拟剂光谱的分类实验表明,经过优化算法训练的分段线性分类器可以用较少的超平面逼近非线性分界面,而且得到更高的识别准确率。     

扩展目标夏克-哈特曼波前传感器的频域迭代算法研究

扩展目标夏克-哈特曼波前传感器(SH-WFS)子图像之间偏移量的计算是影响波前传感精度的关键,通常采用相关算法来实现,并通过抛物线插值达到亚像元精度。子图像间的相对偏移量计算也可采用频域相移量估算的方法进行计算,频域算法还可通过迭代进一步提高算法精度。对频域迭代算法进行了理论分析、仿真和实验研究,结果表明,频域迭代算法在信噪比高于4…1时,具有比抛物线插值法更高的计算精度;在信噪比较低时,与抛物线插值法精度相近。     

基于光幕约束和分段调整的去雾算法

针对雾天条件下成像设备采集到的图像目标主体不清晰,细节信息不明显,可辨识度低等问题,提出了一种基于光幕约束和分段调整的图像去雾算法.首先将大气光幕的基本不等条件映射至正方形及其内切圆平面模型中,得到约束后的初始大气光幕;然后利用饱和度逼近构造粗糙大气光幕,以模糊无用的边缘信息,并通过梯度约束求取边缘代价对初始大气光幕进...     

A method for on-line background subtraction in frequency domain fluorometry

Time-resolved fluorescence determinations on biochemical samples are often complicated by contributions from background. In the time-correlated single-photon counting method, background subtraction is a routine procedure. A limitation of frequency domain fluorometry, however, arises from the difficulty of performing this operation. This limitation has become increasingly significant as frequency domain methods are being applied to evermore complex biological systems using the higher-frequency capabilities of modern instrumentation. We have devised a method for such a correction in the frequency domain, regardless of the complexity of the background decay, based on measurement of the background phasor and subsequent subtraction from the sample phasor. This method is applicable to both lifetime and dynamic polarization measurements, and it can be readily implemented on commercially available frequency domain fluorometers. Decay curves may be accurately recovered from samples containing background contributions ranging from less than 5% to greater than 90% of the total signal intensity.     

基于频谱分析与支持向量机的车型音频识别研究

机动车车型识别是城市道路交通流监测统计的一个重要方面。本文基于频谱分析与支持向量机方法提出一种车型音频识别方法,以1/3倍频程频谱数据作为特征数据,并使用支持向量机方法完成不同车型分类下的车型识别,同时还分析比较了不同训练样本量及不同单个样本数据量大小对识别结果的影响。在将车型细分的情况下,对小汽车、大型公交车、水泥车、摩托车四种车型的样本外识别结果达到96.9%的准确率,验证了方法的有效性。     

Diagnosis of colon cancer using frequency domain fluorescence imaging technique

Early detection and treatment of colon cancer has been associated with better disease prognosis. Conventional and reported optical techniques have limitations in detecting early stages of colon cancer growth. In this paper, a homodyne signal processing assisted frequency domain (FD) fluorescence imaging methodology is proposed for the early diagnosis of colon cancer. Simulated phantom tissues representing the biopsy samples at different stages of colon cancer growth are prepared and used for the imaging study. Selective imaging of healthy and diseased sites simulated in the samples was achieved even for fluorescence emissions having close lifetimes and wavelength values. Possible extension of the methodology for in vivo investigations is also discussed.     

热疗用聚焦脉冲声场的时域和频域特性研究

本文用传递函数的概念导出了凹球面聚焦脉冲声扬的简洁表达式,从数值计算上和实验上研究了这种声场和中轴线上不同位置处声压信号的持续时间和频谱特征,分析了它们对热疗的影响。     

Frequency domain optical tomography using a conjugate gradient method without line search

A conjugate gradient method without line search (CGMWLS) is presented. This method is used to retrieve the local maps of absorption and scattering coefficients inside the tissue-like test medium, with the synthetic data. The forward problem is solved with a discrete-ordinates finite-difference method based on the frequency domain formulation of radiative transfer equation. The inversion results demonstrate that the CGMWLS can retrieve simultaneously the spatial distributions of optical properties inside the medium within a reasonable accuracy, by reducing cross-talk between absorption and scattering coefficients.     

基于OTSU分割和HOG特征的行人检测与跟踪方法

传统的HOG算法针对整幅图像进行行人特征提取,大量的非人窗口计算必然降低检测的准确率和效率。为此,提出一种基于OTSU分割和HOG特征的行人检测与跟踪方法。利用OTSU算法以最佳阈值分割图像,在分割区域的基础上进行Canny边缘检测,通过边缘的对称性计算确定行人候选区,继而采用经PCA方法降维后的HOG特征和隐马尔可夫模型对行人候选区进行检测验证。最后,以确定的行人区域为跟踪窗口,利用CamShift算法跟踪行人。多组实验结果证明,本文方法的行人检测效率和精度均有所提高,跟踪性能稳定、可靠。     

空时多普勒频移域运动小目标的抗干扰探测方法

有源声呐使用多普勒敏感信号可在多普勒频移域分离杂波、混响等干扰和运动目标,但低信干比时强干扰的时间旁瓣泄露和多普勒旁瓣泄露会淹没弱目标。针对该问题,提出了一种基于自适应最小均方(LMS)算法和宽带模糊函数(WAF)的运动小目标抗干扰探测方法。首先在空间多普勒频移域上利用基于LMS的自适应陷波器抑制干扰,然后在快时间多普勒频移域上利用WAF探测小目标。给出了理论推导和仿真分析,进行了水池试验。结果表明,该方法可有效抑制强干扰并能准确估计小目标的时延及多普勒频移,在水池试验中约有13 dB的干扰抑制能力。此外,该方法能够容忍归一化幅度起伏方差小于0.5的干扰起伏,提高了有源声呐对运动小目标的探测能力。     

Fully-automated approach to hippocampus segmentation using a graph-cuts algorithm combined with atlas-based segmentation and morphological opening

The hippocampus has been known to be an important structure as a biomarker for Alzheimer’s disease (AD) and other neurological and psychiatric diseases. However, it requires accurate, robust and reproducible delineation of hippocampal structures. In this study, an automated hippocampal segmentation method based on a graph-cuts algorithm combined with atlas-based segmentation and morphological opening was proposed. First of all, the atlas-based segmentation was applied to define initial hippocampal region for a priori information on graph-cuts. The definition of initial seeds was further elaborated by incorporating estimation of partial volume probabilities at each voxel. Finally, morphological opening was applied to reduce false positive of the result processed by graph-cuts. In the experiments with twenty-seven healthy normal subjects, the proposed method showed more reliable results (similarity index = 0.81 ± 0.03) than the conventional atlas-based segmentation method (0.72 ± 0.04). Also as for segmentation accuracy which is measured in terms of the ratios of false positive and false negative, the proposed method (precision = 0.76 ± 0.04, recall = 0.86 ± 0.05) produced lower ratios than the conventional methods (0.73 ± 0.05, 0.72 ± 0.06) demonstrating its plausibility for accurate, robust and reliable segmentation of hippocampus.     

基于C-V模型的改进快速水平集图像分割法

针对水平集方法计算复杂度高,无法满足实时系统要求的缺陷,提出一种改进的快速水平集算法。该算法对快速水平集算法进行简化,采用单链表表示轮廓曲线。利用C-V模型的二值拟合项来设计曲线演化的速度函数,保留了C-V模型的全局优化特性。还给出了一个基于单链表中轮廓点个数变化的水平集演化终止准则。该算法不仅明显提高了分割速度,且对噪声图像也能实现高效的分割。     

New developments in frequency domain optical tomography. Part I: Forward model and gradient computation

This two part study introduces new developments in frequency domain optical tomography to take into account the collimated source direction in the computation of both the forward and the adjoint models. The solution method is based on the least square finite element method associated to the discrete ordinates method where no empirical stabilization is needed. In this first part of the study, the solution method of the forward model is highlighted with an easy handling of complex boundary condition through a penalization method. Gradient computation from an adjoint method is developed rigorously in a continuous manner through a lagrangian formalism for the deduction of the adjoint equation and the gradient of the objective function. The proposed formulation can be easily generalized to stationary and time domain optical tomography by keeping the same expressions.