近红外光谱无创血糖检测中的生理背景问题

分析了目前直接利用携带复杂组织背景信息的人体近红外光谱进行无创血糖检测方法存在的技术障碍：作为主要信息载体的人体生理背景成分复杂、易变,导致测得的人体近红外光谱无法反映血液中血糖浓度的变化信息,因此,其检测精度很难达到临床应用的要求。提出一种新方法,即利用血流容积一直在变而人体组织背景和血液成分含量短时间不变的事实,消除人体组织背景干扰,获得只反映血液信息的＂纯净＂光谱,将近红外无创血糖检测技术的重点问题转移到如何得到具有较高信噪比的光谱数据,为该技术的临床应用提供了可能性。     

一种改进的小波基线校正方法的研究

通过对小波变换基线校正中最佳分解尺度方法的研究,提出了一种新的基于小波变换的最佳分解尺度确定方法,不但有效地提高了基线校正效果,而且具有简单、快速的优点.将该方法应用于血糖光谱数据预处理中进行基线校正,取得了较好的效果.通过人体口服葡萄糖耐量试验(OGTT)得到人体无创检测近红外光谱和对应血糖浓度值,然后采用该方法对上述光谱进行基线校正并建立多元回归模型,采用交互验证的方式对模型及基线校正的效果进行了评价.实验结果表明,血糖浓度预测值和参考值间的相关系数为0.75,预测均方根误差(RMSEP)为1.36 mmol/L,与原始光谱预测结果和其他小波分解尺度下的预测结果相比,RMSEP降低了将近39%,相关系数提高了0.64,预测精度得到较大幅度提高.     

血糖近红外定量分析方法研究

为解决传统有创检测血糖的问题,文中对近红外光谱无创血糖检测技术进行了研究。使用ABB公司的TALYS ASP531近红外光谱仪测得葡萄糖溶液的光谱,建模波段选取5 500~6 500 cm ^-1,使用偏最小二乘法建立光谱数据与对应浓度之间的校正模型。实验结果表明,校正模型的决定系数达到了0.983 99,校正均方根偏差为1.023 26 mmol·L ^-1,误差分布为-1.3~1.7 mmol·L ^-1,证明了近红外无创检测人体血糖浓度的可行性,为近红外血糖检测的研究提供了可靠的依据。     

利用可见/近红外光谱测定小麦叶面积指数的改进研究

改进了小麦叶面积指数的可见/近红外光谱测定模型。以不同方法实现了小麦冠层反射光谱的预处理,并采用偏最小二乘回归算法（PLS）建立小麦叶面积指数估测模型对其进行比较分析,发现小波除噪结合一阶导数能最有效地消除原始光谱的噪声与背景信息,此时PLS模型校正集与预测集R2分别为0.849与0.835。为进一步优化模型,对经一阶导数结合小波除噪后的光谱采用主成分分析法（PCA）降维,以前4个主成分（含原始光谱84.867%特征信息）为输入变量,采用小二乘支撑向量机回归算法（LS-SVR）建立了小麦叶面积指数估测模型,其校正集与预测集R2分别达0.905与0.883,具有比PLS算法更高的精度。结果表明:以小波除噪结合一阶导数去除小麦冠层反射光谱中的土壤背景信息以提高模型精度是可行的,且LS-SVR是建模的优选方法。     

近红外光谱结合小波变换-径向基神经网络用于奶粉蛋白质与脂肪含量的测定

应用近红外光谱分析技术结合化学计量学方法,建立了奶粉脂肪和蛋白质含量测定的化学计量学建模新方法.首先采用Kernard-Stone法对校正集样本和预测集样本进行分类,然后利用小波变换滤波技术对样品的近红外光谱进行压缩去噪处理,结合滤波后重构光谱信号建立脂肪和蛋白质的径向基神经网络回归模型,并分别对径向基网络的扩散常数spread值及小波变换中的小波基与压缩尺度三个参数进行了详细的讨论.脂肪模型在小波基为db2及小波尺度为4、spread值为3.5时的预测模型精度最好;蛋白质模型在小波基为db8及小波尺度为4、spread值为6时,预测模型精度最好.所建立的脂肪和蛋白质校正模型的预测集相关系数(Rp)分别为0.990和0.994,预测均方根误差分别为0.007与0.004.预测结果表明,RBF网络结合小波变换进行建模预测,模型简单、稳健且精度较好,该方法适合奶粉脂肪和蛋白质含量的快速、无损测定.     

基于近红外光谱技术乳制品品种快速无损鉴别

研究了使用近红外透射光谱法对四种乳制品品种进行快速鉴别的方法.首先对样品光谱进行多元散射校正等预处理,然后对处理后的光谱进行小波基为db3,分解尺度为6的小波压缩,最后以压缩后光谱数据作为输入变量,分别采用BP-ANN、RBF-ANN和SVM-ANN三种人工神经网络建立品种鉴别模型.结果显示,当光谱预处理方法为多元散射校正加S-G平滑加小波变换(MSC +S-G +WT)时,三种网络对样品的鉴别率均达到了100％.所以近红外光谱技术可以实现对乳制品品种的快速无损鉴别.     

结合光谱和尺度特征的高分辨率图像边缘检测算法

高分辨率遥感图像具有高度细节化的多尺度表达能力,在有效表达地物边缘信息的同时,目标内部几何细节常以噪声的形式出现.提出将光谱相异性和小波变换相结合的边缘特征检测算法,克服了小波变换导致的边缘变形,并能够有效抑制噪声.根据光谱角原理定义归一化光谱相异性模型,并与二进小波变换结合,同时利用梯度方向余弦值对各个波段的梯度幅值加权,最后根据向量场模型计算多光谱图像的梯度幅值和梯度方向,细化后获取由细到粗的多层次边缘特征.实验结果与小波变换和传统检测算子的检测结果相比,表明该算法利用光谱相异性信息增强边缘响应强度,保证了所有尺度下获取的边缘轮廓不失真,边缘点定位准确;加权处理突出了多波段梯度主方向信息,也有效抑制了高分辨率图像上目标内部精细几何细节形成的噪声.     

苹果近红外光谱的预处理

原始近红外光谱数据含有大量的噪声信号和较大的数据量,所以在进行光谱数据分析之前对光谱数据进行预处理是非常必要的。近红外光谱数据的预处理主要有两个任务,一是降噪,提高模型的稳健性和预测结果的准确性;二是数据压缩,以便于数据的存储,提高建模速度。传统的近红外光谱数据预处理方法各有局限,很难在这两方面都得到令人满意的效果。将小波分析用于苹果近红外光谱数据的预处理,并选取峰值信噪比(Peak Signal to Noise Ratio, PSNR)和归一化相关系数(Normalized Correlation, NC)作为评价指标。与常用的Savitzky-Golay平滑滤波和多元散射校正相比,小波方法不仅能有效地实现数据压缩,而且在噪声去除和光谱细节保持等方面都具有优势。}     

基于AR模型的小波变换在脑电信号消噪中的应用

提出了基于自回归模型(ARM)与小波变换的脑电信号分析方法,并利用他来消除脑电信号中的噪声干扰。小波变换是一种多分辨率的时间尺度分析方法,他能够将信号划分为不同频段的子带信号。根据小波变换的这一特性,对采样获得的脑电信号进行各尺度分解及消噪分析,并给出了各尺度分解结果及消噪结果。利用小波变换能有效去除脑电信号中的噪声干扰。     

基于多尺度阈值技术的小波去噪

多尺度阈值技术的小波去噪是根据小波变换下信号和噪声在多尺度空间中的不同特性,用软阈值函数确定的阈值对不同尺度上的小波系数进行处理,进而达到去除噪声的目的。运用这种方法对信号进行处理时去噪效果良好,并且能够很好地保留原信号的细节信息。     

天空背景下一维红外小目标检测技术研究

针对小波去噪法在处理天空背景下一维红外小目标时不能有效去除包含在近似系数中的高频噪声的问题,提出了首先通过Top-Hat算法对近似系数进行背景对消,然后采用改进阈值函数对细节系数进行量化的检测方法。数字仿真结果表明,新方法能有效地去除噪声干扰,提高信噪比。     

Extraction of periodic multivariate signals: mapping of voltage-dependent dye fluorescence in the mouse heart

In many experimental circumstances, heart dynamics are, to a good approximation, periodic. For this reason, it makes sense to use high-resolution methods in the frequency domain to visualize the spectrum of imaging data of the heart and to estimate the deterministic signal content and extract the periodic signal from background noise in experimental data. In this paper, we describe the first application of a new method that we call cardiac rhythm analysis which uses a combination of principal component analysis and multitaper harmonic analysis to extract periodic, deterministic signals from high-resolution imaging data of cardiac electrical activity, We show that this method significantly increases the signal-to-noise ratio of our recordings, allowing for better visualization of signal dynamics and more accurate quantification of the properties of electrical conduction. We visualize the spectra of three cardiac data sets of mouse hearts exhibiting sinus rhythm, paced rhythm and monomorphic tachycardia. Then, for pedagogical purposes, we investigate the tachycardia more closely, demonstrating the presence of two distinct periodicities in the re-entrant tachycardia. Analysis of the tachycardia shows that cardiac rhythm analysis not only allows for better visualization of electrical activity, but also provides new opportunities to study multiple periodicities in signal dynamics.     

基于小波变换的话务量预测

本文将小波变换应用到话务量的预测中,利用小波分解法将非平稳时间序列的GSM话音话务量分解为多个细节信号分量和逼近信号分量。对细节信号采用AR模型或者余弦逼近进行拟合建模,对逼近信号采用多项式拟合和AR模型相结合的方式建模。利用某运营商2009年1月~2013年7月每月的博彩日话音话务作为检验序列集,前50个月的数据用来建模拟合,最后5个月数据作为预测比较,发现拟合相关度为0.991,预测平均绝对误差为0.029,预测结果比单纯使用曲线拟合要好。     

On the noise properties of linear and nonlinear quantum-dot semiconductor optical amplifiers: the impact of inhomogeneously broadened gain and fast carrier dynamics

We present a detailed analytical model describing the noise properties of quantum-dot (QD) optical amplifiers operating in the linear and saturated regimes. We describe the dependence of the optical noise on the main physical parameters characterizing the QD gain medium as well as on operating conditions. The optical noise at the amplifier output shows a broad-band spectrum with an incoherent spectral hole due to the gain inhomogeneity. A coherent spectral dip stemming from noise-signal nonlinear interactions is superimposed on that broad-band spectrum. The broad-band incoherent component is also calculated using an approximate model which makes use of an equivalent inhomogeneous population inversion factor. The validity of the approximation is examined in detail. We also calculate the electrical relative intensity noise and observe a spectral hole corresponding to the spectral shape of the optical noise. The most important characteristics of the optical and electrical noise spectra are determined by the degree of inhomogeneous broadening and by the fast carrier dynamics of QD amplifiers. The fast dynamics causes a very wide noise spectral hole which has important potential consequences for detection of fast data and for all optical signal processing.     

基于方向滤波器组的指纹图像预处理

提出一种变换域指纹图像预处理增强方法。应用小波变换分离出低频信号和高频信号,对高频信号送入方向性滤波器组进行方向性子带分解,对各方向子带图像分块,根据块能量分布情况对子带图像重新加权处理。子带图像重建出高频信号加上原低频信号得到预处理结果,达到抑制噪声,增强特征目的。实验表明算法对指纹图像处理效果良好。     

Digital Coding of Color Video Signals--A Review

This paper reviews the field of the efficient coding of color television signals. Because this paper is perhaps the first review on this topic, some background is given in the areas of colorimetry, visual perception of color and color television systems. We assume that the reader has some familiarity with luminance encoding techniques. Coding techniques themselves are divided into two broad groups: component coding methods in which each component (usually three) is coded separately, and composite coding methods in which the composite television signal with its "color" modulated subcarrier is processed as a single entity. Both approaches are covered in detail. The field is still growing, pushed primarily by the desire in the television area to find digital coding standards accepted by both broadcasters and carriers and suitable for use with NTSC, PAL and SECAM television systems. We discuss this aspect by comparing composite and component coding methods.     

A lens antenna array with adaptive optical processing

This paper presents an X band smart antenna array in which adaptive processing of the received signals is performed by dynamic holographic optical circuitry. The optical circuitry adaptively extracts the principal component of the received signal space, that is the strongest first-order independent temporal component of the ensemble of received signals. The adaptive receiver system can be used, for example, to mitigate multipath interference effects and can separate one received signal from another even though their power spectra may entirely overlap. A prototype two-channel system is designed to fit in a standard-size briefcase and consume less than 50 W of power. The input to the system are modulated waves with a carrier in X band and the output is an electronic demodulated signal. Three major components of this system are described in detail: (1) the quasi-optical lens antenna array front end with angle-of-arrival preprocessing and downconversion, (2) the two-channel electrooptic modulation and optical carrier suppression stage, and (3) the smart optical processor (auto-tuning filter). Component and end-to-end system measurements give quantitative indicators for the usefulness of optical processing in wireless communications     

复杂背景下基于分割逼近法的抠像技术研究

论述了复杂背景下人物的抠像问题,提出一种新的抠像方法--分割逼近算法.首先使用背景更新模型建立了差分背景,然后用本文提出的分割逼近法对图像进行粗分和细分两次分割,建立了人物图像边缘连通体,最后运用Laplacian算子对勾勒出的人物轮廓提取边缘,并用替换原图像的方法实现人物的抠像.     

On classification of multispectral infrared image data

A detector is proposed that is based on a model in which the signal components consist of radiant thermal energy from either the small target or the intense, highly structured background. The resulting statistic is effective in enhancing the target and suppressing cluttered background. Estimation of the system parameters based on stochastic approximation techniques is presented. Simulation results demonstrate the practicality of the proposed detector.     

Effect of bit-threshold errors on the harmonic and intermodulationperformance of successive approximation A/D converters

A simple approach is described for predicting the harmonic and intermodulation performance of a real successive approximation analog-to-digital converter (ADC) with bit-threshold errors. The real ADC characteristic is split into an ideal staircase characteristic, representing the ideal ADC characteristic, minus a train of pulses, representing the bit-threshold errors. Series expressions are obtained for the output spectra of the real ADC excited by a multicarrier input signal with arbitrary amplitudes. The special case of two equal-amplitude carriers is considered in detail, and analytical expressions are obtained for the amplitudes of the harmonic and intermodulation products in terms of the individual bit-threshold errors. The feasibility of minimizing the amplitudes of the intermodulation products is explored