Detection of laryngeal function using speech andelectroglottographic data

The purpose of this research was to develop quantitative measures for the assessment of laryngeal function using speech and electroglottographic (EGG) data. We developed two procedures for the detection of laryngeal pathology: 1) a spectral distortion measure using pitch synchronous and asynchronous methods with linear predictive coding (LPC) vectors and vector quantization (VQ) and 2) analysis of the EGG signal using time interval and amplitude difference measures. The VQ procedure was conjectured to offer the possibility of circumventing the need to estimate the glottal volume velocity wave-form by inverse filtering techniques. The EGG procedure was to evaluate data that was "nearly" a direct measure of vocal fold vibratory motion and thus was conjectured to offer the potential for providing an excellent assessment of laryngeal function. A threshold based procedure gave 75.9 and 69.0% probability of pathological detection using procedures 1) and 2), respectively, for 29 patients with pathological voices and 52 normal subjects. The false alarm probability was 9.6% for the normal subjects.     

Bounds on the pairwise error probability of coded DS/SSMAcommunication systems in Rayleigh fading channels

Arbitrarily tight upper and lower bounds on the pairwise error probability (PEP) of a trellis-coded or convolutional-coded direct-sequence spread-spectrum multiple-access (DS/SSMA) communication system over a Rayleigh fading channel are derived. A new set of probability density functions (PDFs) and cumulative distribution functions (CDFs) of the multiple-access interference (MAI) statistic is derived, and a modified bounding technique is proposed to obtain the bounds. The upper bounds and lower bounds together specify the accuracy of the resulting estimation of the PEP, and give an indication of the system error performance. Several suboptimum decoding schemes are proposed and their performances are compared to that of the optimum decoding scheme by the average pairwise error probability (APEP) values. The approach can be used to accurately study the multiple-access capability of the coded DS/SSMA system without numerical integrations     

基于基元段特征和GMM的源-目标说话人F0～t转换

基音轨迹F_0～t转换是实现高质量源-目标说话人声音转换的重要组成部分。本文给出了一种与说话内容无关的F_0～t转换方法,为了在与文本无关的前提下,提取出尽可能反映说话人个性特征的基音起伏的较长时的信息,本文采用了从n个短时帧组成的基元段提取特征矢量,并以基元段矢量为单元进行转换,采用了基于高斯混合模型(GMM)的概率加权转换算法使每个特征矢量的转换规则是由多个类规则的线性加权组合得到的,从而提高了转换精度,同时还解决了一般分类器中处于类边界数据的分类错误。实验表明,基于基元段特征矢量和GMM的转换方法具有很好的效果。     

概率框架下多特征显著性检测算法

显著性检测是计算机视觉的一项基础问题,广泛地用于注视点预测、目标检测、场景分类等视觉任务当中.为提升多特征条件下图像的显著性检测精度,以显著图的联合概率分布为基础,结合先验知识,设计一种概率框架下的多特征显著性检测算法.首先分析了单一特征显著性检测的潜在缺陷,继而推导出多特征下显著图的联合概率分布;然后根据显著图的稀有性,稀疏性,紧凑性与中心先验推导出显著图的先验分布,并使用正态分布假设简化了显著图的条件分布;随后根据显著图的联合概率分布得到其极大后验估计,并基于多阈值假设构建了分布参数的有监督学习模型.数据集实验表明：相比于精度最高的单一特征显著性检测方法,多特征算法在有监督和启发式方法下的平均误差降低了6.98%和6.81%,平均F-measure提高了1.19%和1.16%;单幅图像的多特征融合耗时仅为11.8ms.算法精度较高,实时性好,且可根据不同任务选择所需的特征类别与先验信息,能够满足多特征显著性检测的性能要求.     

Error Probability Distribution and Density Functions for Weibull Fading Channels With and Without Diversity Combining

In this letter, a detailed theoretical analysis of probability distribution and density functions of probability of error in a wireless system is considered. Closed form expressions for distribution and density functions of the probability of error are derived for Weibull fading channels for the cases of (i) No Diversity (ND), (ii) Selection Combining (SC) diversity, and (iii) Switch and Stay Combining (SSC) diversity. Numerical results are plotted and discussed in detail for the various cases.     

基于深度自编码-高斯混合模型的视频异常检测方法

由于异常定义的模糊性和真实数据的复杂性,视频异常检测是智能视频监控中最具挑战性的问题之一。基于自动编码器(AE)的帧重建（当前或未来帧）是一种流行的视频异常检测方法。使用在正常数据上训练的模型,异常场景的重建误差通常比正常场景的重建误差大得多。但是,这类方法忽略了正常数据本身的内部结构,效率较低。基于此,提出了一种深度自动编码高斯混合模型(DAGMM)。首先利用深度自动编码器获得输入视频片段的生成低维表示和重构误差,并将其进一步输入高斯混合模型(GMM)。而估计网络则通过高斯混合模型预测能量概率,然后通过能量密度概率判断异常。DAGMM以端到端的方式同时联合优化深度自动编码器和GMM的参数,能够平衡自动编码重建、低维表示的密度估计和正则化,泛化能力强。在两个公共基准数据集上的实验结果表明,DAGMM达到了现有最高技术发展水平,在UCSD Ped2和ShanghaiTech两个数据集上分别取得了95.7%和72.9%的帧级AUC。     

Automated Osborn Wave Detection System Based on Wavelet Features and Neural Network

Automated Osborn wave detection system featuring the sensitivity of 94.63% and classification accuracy of 94.58% for the notch and slur types of waves in the cardiac signal has been developed. The quasi-matched wavelet filtering method and the method of principal components were applied for extraction and formation of feature vectors representing the input data of classifier. The error feedforward neural network with topology of a multilayer perceptron was used as a classifier. Signal samplings built on information from the PhysioNet open database of medical signals were used for training, testing and validation of neural network. This study involved the use of 12-lead electrocardiograms of 60 healthy patients aged 17–87. These electrocardiograms formed the basis for creating a database of 14832 signals (9888 with Osborn wave signals of two types and 4944 signals without pathological findings). The proposed approach ensured the classification accuracy exceeding the accuracy of existing techniques.     

Error probabilities in binary angle modulation

It is shown that many signal-detection problems involving binary phase and frequency modulation can be solved by considering the problem as the convolution of two independent phase distributions. When the two distributions are similar and each represents the probability density function of the phase of a sinusoid in noise, the probability of error is described in terms of modified Bessel functions by the expression obtained recently by Jain and Blachman. The probability of error can also be expressed in terms of Rice'sIefunction. By appropriate definition of the signal-to-noise ratios, this general expression can be used to determine the error probability in a number of cases of practical interest, such as detection of a hard-limited phase-shift keyed (PSK) signal, PSK detection with noisy reference, DPSK detection, frequency-shift keying (FSK) detection, binary FM with discriminator detection, and binary pulse-position modulation (PPM).     

A Trivariate Nakagami-m Distribution with Arbitrary Covariance Matrix and Applications to Generalized-Selection Diversity Receivers

This paper deals with a trivariate Nakagami-m distribution derived from the diagonal elements of a Wishart matrix. For this distribution, infinite series representations for its probability density and cumulative distribution functions are derived having an arbitrary covariance matrix and integer-order fading parameters. Moreover, upper bounds on the error resulting from truncating the infinite series are obtained. Based on the derived formulas, the performance of triple-branch generalized selection combining (GSC) receivers is analyzed. For this type of receivers, the outage and the average bit error probability for a variety of modulation schemes are analytically obtained. The performance of GSC receivers is compared to that of conventional selection and maximal-ratio diversity schemes. In order to check the accuracy and convergence of the derived formulas, various performance evaluation results are presented and compared to equivalent simulation ones.     

Performance and Optimization of Switched Diversity Systems for the Detection of Signals with Rayleigh Fading

The performance and optimization of switched diversity systems are considered. First, the one-dimensional distribution and probability density functions of the envelope of the received signal are obtained for three different switching strategies. This information is used to obtain the average probability of bit error for the case of non-coherent detection of binary FSK signals with Rayleigh fading envelopes and additive white Gaussian noise. The optimization of two of these switching strategies is then considered, and it is shown that by proper selection of switching thresholds, the average probability of bit error during detection can be minimized. It is also shown that these optimized switching strategies yield a significant improvement in performance over non-diversity systems and can approach the performance of more complex receivers such as maximal ratio combining. Computer simulations of switched diversity systems using a practical field model are used to verify the analysis.     

The Influence of Cycle Slipping on the Error Probability of a PAM Receiver

In this paper, we evaluate the error probability of a PAM receiver, which derives symbol timing from the received waveform by means of a synchronizer. We point out that the conventional way (using the distribution of the synchronizer's modulo-2π- reduced phase error) of evaluating this error probability does not take into account the effect of the synchronizer's cycle slips. We present a correct expression for the error probability, which uses the distribution of the so-called renewal phase. Also, we derive simple but accurate approximations for both the conventional and the correct error probability, which clearly show the influence of the characteristics of the PAM receiver. These approximate expressions indicate that, for a decreasing additive noise level at the receiver input, the conventional error probability becomes vanishingly small, whereas the correct error probability approaches a nonzero value, which is inversely proportional to the mean time between slips. Hence, cycle slipping imposes a (sometimes rather large) lower limit on the attainable error probability.     

Recursive Bayesian Algorithm with Covariance Resetting for Identification of Box–Jenkins Systems with Non-uniformly Sampled Input Data

To identify the Box–Jenkins systems with non-uniformly sampled input data, a recursive Bayesian algorithm with covariance resetting was proposed in this paper. Considering the prior probability density functions of parameters and the observed input–output data, the parameters were estimated by maximizing the posterior probability distribution function. During the estimation, the variance of the noise was taken as a weighting factor, and the proposed algorithm was formulated as a weighted least squares. As a result, the accuracy of the estimates increased. Meanwhile, a modified covariance resetting strategy was integrated into the algorithm to improve the convergence rate, and the convergence of the algorithm was also analyzed. A simulation example was applied to validate the proposed algorithm.     

Worst-case power-constrained noise for binary-input channels

Additive noise channels with binary-valued inputs and real-valued outputs are considered. The maximum error probability and the minimum channel capacity achieved by any power-constrained noise distribution are obtained. A general framework which applies to a variety of performance measures shows that the least-favorable noise distribution is, in general, a mixture of two lattice probability mass functions. The framework holds for m-ary input constellations on finite-dimensional lattices.<>     

Fault Detection and Diagnosis for General Stochastic Systems Using B-Spline Expansions and Nonlinear Filters

This paper presents a new fault detection and diagnosis (FDD) algorithm for general stochastic systems. Different from the classical FDD design, the distribution of system output is supposed to be measured rather than the output signal itself. The task of such an FDD algorithm design is to use the measured output probability density functions (PDFs) and the input of the system to construct a stable filter-based residual generator such that the fault can be detected and diagnosed. For this purpose, square root B-spline expansions are applied to model the output PDFs and the concerned problem is transformed into a nonlinear FDD algorithm design subjected to a nonlinear weight dynamical system. A linear matrix inequality based solution is presented such that the estimation error system is stable and the fault can be detected through a threshold. Moreover, an adaptive fault diagnosis method is also provided to estimate the size of the fault. Simulations are provided to show the efficiency of the proposed approach.     

Minimum Classification Error Training Incorporating Automatic Loss Smoothness Determination

Minimum Classification Error (MCE) training, which has been widely used as one of the recent standards of discriminative training for classifiers, is characterized by a smooth sigmoidal-form classification error count loss. The smoothness of this loss function effectively increases training robustness to unseen samples, well approximates the ultimate, minimum classification error probability status, and leads to accurate classification over unseen samples. However, few rational methods have been developed for controlling the smoothness, which is often determined through many repetitions of the experimental setting; this empirical approach has been a disincentive to the increased popularization of MCE training. To alleviate this long-standing problem, we propose a new MCE training method that automatically determines loss smoothness. The proposed method is based on Parzen-estimation-based MCE re-formalization, and the loss smoothness degree is determined so that Parzen distribution can be an accurate approximation to the unknown true distribution, whose positive-domain integration corresponds to classification error probability, in one-dimensional misclassification measure space. Through systematic experiments, we show that the proposed method efficiently yields a classification accuracy that nearly matches the best accuracy obtained by the conventional, trial-and-error-mode repetitions of smoothness setting.     

基于雷达测量的多目标联合检测、跟踪与分类方法

利用雷达测量中的目标速度、加速度等属性信息, 基于跳转马尔科夫系统模型高斯混合概率假设密度滤波算法, 提出了一种多目标联合检测、跟踪与分类方法.该方法在进行雷达多目标测量信息处理的多模型混合高斯概率假设密度滤波过程中, 对各高斯项编号, 进行航迹提取, 在滤波处理的同时形成带有航迹编号的明确航迹, 并进行航迹管理; 同时, 根据目标运动模型, 联合利用目标加速度控制输入与速度估计进行多目标分类.仿真试验验证了该方法能够在检测、跟踪的同时, 对目标航迹进行有效类型识别.     

一种基带GMSK信号相关器及其输出概率分布

该文提出一种基带GMSK信号相关器,并从GMSK解调信号的相位概率分布函数以及独立同分布随机变量和的概率分布函数出发,给出了该相关器的自相关峰和互相关峰的概率分布函数。从概率论与数理统计出发,推导了一种统一的数字相关器(数字匹配滤波器)的自相关峰和互相关峰的概率分布函数。数值计算结果表明,在相同条件下,该基带GMSK信号相关器的误检概率比数字相关器的低近一个数量级。     

Unified analysis of error performance for binary signalling over Rayleigh fading channels

Using the unified expression as devised by Wojner, an analytical framework is presented to determine error performance of all binary modulation schemes employing coherent or non-coherent detection and operating over a slow flat Rayleigh fading channel with different diversity combining. A class of generalised expressions for the bit error probability has been obtained following the probability density function or the cumulative distribution function approach.     

Classification of voiceband data signals using the constellationmagnitude

We discuss the real time classification of high speed (⩾4800 b/s) voiceband data signals using blind or self-recovering equalization techniques. After equalization, the signals are classified, by means of their constellation magnitude, using a new classification scheme which computes the probability that the unknown test signal belongs to each of several predetermined classes. Heuristic rules are applied to these probability functions to perform the actual signal detection. Even in the presence of added channel impairments, the use of this new approach resulted in an overall classification accuracy of greater than 98%, measured using 5100 test segments collected from 6 classes of signals produced by V.29, V.32, and V.33 modems. Classification times range from median values of 0.090 s and 0.219 s for the V.29 modem at 4800 and 7200 b/s, respectively, up to a worse case time of 3.33 s for the 9600 b/s default class, chosen only if no other signal is detected within a prescribed amount of time