一种新型的非参数模型及其在系统辨识中的应用

本文给出了一个求解广义正交多项式的微分运算矩阵的新方法，应用对连续线性系统的脉冲响应函数进行正交逼近的方法来讨论脉冲响应函数的实现问题，得到了一类新型的非参数模型，并导出了利用该模型来辨识连续线性系统的脉冲响应函数的算法，最后给出了例子证实本文所给方法的有效性。     

基于脉冲响应函数的正交小波变换系数的故障检测方法

提出了一种基于连续系统脉冲响应函数的正交小波变换系统的故障检测方法。该方法无需对象的数学模型,具有计算量小、故障判决过程简单、实时性较好等优点。     

Discrete-time frequency response identification method for processes with final cyclic-steady-state

A new non-parametric process identification method is proposed to obtain the discrete-time frequency response model from the process input and output data. The existing discrete-time Fourier transform approach can be applied to only the case that the initial part and the final part of the process data are zero-steady-state to estimate perfect frequency response data without modeling errors. The proposed method using a new transform can estimate the exact frequency response model from more various process excitation cases including initial-steady-state/final-steady-state and initial-steady-state/final-cyclic-steady-state. It can estimate exact frequency response model because no approximations are used in developing the proposed algorithm. Also, the proposed method can still provide exact model even in the case of static disturbances and sinusoidal disturbances of which the frequencies are multiples of the cyclic-steady-state frequency.     

Frequency Response Estimation from Impulse or Step‐like Response by Virtual Experiments

This paper considers the problem of estimating the frequency response of an unknown plant from impulse or step response measurements, possibly with non‐ideal step signals. The proposed method is direct (no model identification of the plant is needed) and it can be applied using a single set of data generated by the plant, with no need for specific experiments nor iterations. The frequency response of the plant, at a desired frequency, is estimated by a virtual experiment generating a filtered version of the data set. Numerical simulations show the effectiveness of the method.     

On continued fraction expansion of real roots of polynomial systems, complexity and condition numbers

We elaborate on a correspondence between the coefficients of a multivariate polynomial represented in the Bernstein basis and in a tensor-monomial basis, which leads to homography representations of polynomial functions that use only integer arithmetic (in contrast to the Bernstein basis) and are feasible over unbounded regions. Then, we study an algorithm to split this representation and obtain a subdivision scheme for the domain of multivariate polynomial functions. This implies a new algorithm for real root isolation, MCF, that generalizes the Continued Fraction (CF) algorithm of univariate polynomials.A partial extension of Vincent’s Theorem for multivariate polynomials is presented, which allows us to prove the termination of the algorithm. Bounding functions, projection and preconditioning are employed to speed up the scheme. The resulting isolation boxes have optimized rational coordinates, corresponding to the first terms of the continued fraction expansion of the real roots. Finally, we present new complexity bounds for a simplified version of the algorithm in the bit complexity model, and also bounds in the real RAM model for a family of subdivision algorithms in terms of the real condition number of the system. Examples computed with our C++ implementation illustrate the practical aspects of our method.     

Modeling of transient bending wave in an infinite plate and its coupling to arbitrary shaped piezoelements

Estimating the location and energy of impacts is of primary importance for assessing the condition of structures. Particularly, such estimation can be easily obtained from the energy flow in the structures, which is usually derived from the Poynting vector. In order to measure the Poynting vector in a thin plate using piezoelements bonded on the plate, an analytical formulation of the impulse response in thin infinite plates is presented. The knowledge of the impulse response of any linear time invariant (LTI) system is precious information for the determination of its behavior under arbitrary inputs. When dealing with propagation, and especially mechanical wave propagation, a common approach consists in using numerical methods that are often time-consuming, especially for multi-coupled systems. This paper proposes a new approach for modeling the impulse response of an infinite plate with surface-bonded piezoelectric elements. The proposed analytical formulation allows bypassing numerical analysis drawbacks, in particular instabilities occurring at high frequencies, case-dependent systems and computational requirements, while giving the response for any time and space domain values using a simple convolution. The proposed model relies on flexural wave decomposition over the spatial frequency domain and corresponds to a time generalization of the angular spectrum theory, thus introducing flexural wave propagation as a time-varying spatial filter. Once the impulse is know in the spatial frequency domain, the inverse Fourier transform is applied and leads to the impulse response in the physical domain. From this model, an analytical expression of the impulse voltage response of the piezoelectric transducers and the Poynting vector can be derived quite easily. The predicted impulse response is then compared to FEM simulation results and experimental measurements in order to assess the model.     

基于偏置继电反馈的一种新的频域辨识方法

提出一种新的基于偏置继电反馈的频域辨识算法,能够对包括迟延在内的SISO系统的传递函数参数进行在线辨识。首先利用偏置继电反馈辨识过程的频率响应,一次继电测试可以辨识过程Nyquist曲线上的多点。接着利用最小二乘法估计过程传递函数参数的初值,最后应用非线性共轭梯度法准确的辨识传递函数参数。该方法辨识精度高,同时可以有效避免大多继电反馈辨识方法测试时间过长及只能辨识固定的低阶模型的缺点。仿真结果验证了算法的有效性与准确性。     

随机噪声框架下非伪频率响应的计算方法

推导了当系统输入输出实验数据精确已知时,由未建模动态产生的频率响应集合解析表达式.采用序贯蒙特卡洛方法产生随机噪声序列,由这些噪声序列计算出相应的非伪频率响应集.提出一个找到包含所有频率响应且有最小半径的圆的数值优化算法,通过该算法,得到一个以指定概率包含控制对象非伪频率响应的最小集合.最后给出了仿真实例,验证了整个算法的有效性.     

白噪声驱动的线性系统动态特性测量方法

冲激响应反应了线性系统的内部动态特性。该文详细介绍了系统冲激响应辨识的时域互相关技术。它用白噪声对系统进行驱动,将离散时间线性系统转化为时间序列的ARMA模型,并将基于MA模型的系统辨识问题转化为一个优化问题,以最小二乘标准求解。求解过程利用采样样本估计相关函数,构造对称的Toeplitz矩阵方程,再用迭代的Levinson-Durbin算法解该矩阵方程。最后,文章给出了一个线性系统的测量实例。实验结果表明,该方法能准确的估计线性系统的冲激响应。     

Fractional-order impulse response of the respiratory system

This paper presents the measurement, frequency-response modeling and identification, and the corresponding impulse time response of the human respiratory impedance and admittance. The investigated adult patient groups were healthy, diagnosed with chronic obstructive pulmonary disease and kyphoscoliosis, respectively. The investigated children patient groups were healthy, diagnosed with asthma and cystic fibrosis, respectively. Fractional order (FO) models are identified on the measured impedance to quantify the respiratory mechanical properties. Two methods are presented for obtaining and simulating the time-domain impulse response from FO models of the respiratory admittance: (i) the classical pole-zero interpolation proposed by Oustaloup in the early 90s, and (ii) the inverse discrete Fourier Transform (DFT). The results of the identified FO models for the respiratory admittance are presented by means of their average values for each group of patients. Consequently, the impulse time response calculated from the frequency response of the averaged FO models is given by means of the two methods mentioned above. Our results indicate that both methods provide similar impulse response data. However, we suggest that the inverse DFT is a more suitable alternative to the high order transfer functions obtained using the classical Oustaloup filter. Additionally, a power law model is fitted on the impulse response data, emphasizing the intrinsic fractal dynamics of the respiratory system.     

一种基于DCT的频域运动估值算法

用软件实现在低码率信道上的实时视频编解码是图象压缩领域的研究热点，本文就此提出了一种基于离散余弦变换（DCT）系数的频域块匹配法，通过引入阈值矩阵，降低了运动估值的运算量，提高了搜索效率。计算机仿真表明，该算法可在PC机上用软件实时实现，具有较好的实用价值。     

Approximation of discrete linear systems based on a dual criterion

A new method for approximating scalar discrete linear systems is described. The method can be classified as a time domain/frequency domain technique. Two reduced order models are obtained. One is obtained when the time domain specifications are in terms of the impulse response of the error. The other is derived when the time domain specifications are in terms of the step response of the error. The technique utilizes the input normal system representation to determine the order and the denominator polynomial of the reduced order model.     

基于快速m序列变换的线性网络冲激响应测量算法

为探测强噪声干扰下的复杂系统性能，实现了快速m序列变换（Fast m-sequence transform，FMT）算法。在对一般m序列和互补特征m序列的相关函数等属性分析的基础上，探讨了FMT的内部机制及排列矩阵的构造原理，并针对不同的本原多项式，找出了排列矩阵对应的下标索引向量。仿真实验表明基于FMT的冲激响应测量法具有很高的准确性和抗噪声性能。     

基于第二代轮廓波变换的消噪方法

提出一种基于第二代轮廓波变换的消噪方法,该方法在阈值选取上采用基于自适应冲激响应的多尺度多方向阈值,在硬阈值处理之后采用自适应维纳滤波。该方法充分利用了第二代轮廓波变换在频域和时域的局部化、多尺度、多方向和低冗余特征;能够利用更多较小的轮廓波变换系数以减轻轮廓波硬阈值消噪带来的纹理效应,并能更好适应人类的视觉特征。实验结果表明提出的消噪方法在峰值信噪比和视觉效果上都优于第一代轮廓波和小波的消噪效果。     

基于PEM的三自由度直升机模型辨识

针对三自由度(3-DOF)直升机平台的特点,提出了一种基于预测误差法(PEM)的模型频域辨识方法,建立了机理模型,运用扫频技术得到巡航飞行状态直升机3个通道的输入-输出数据;分析了偏相干函数和复合窗函数,通过PEM进行了模型的频域辨识,得到了状态空间方程的待辨识参数和直升机的参数化模型.通过时域飞行和模型预测响应的对比,验证了该模型的准确性和该辨识方法的有效性.     

光学图象几何畸变的快速校正算法

实际光学镜头所成的图象难免会有几何变形，多项式坐标变换法是进行几何修正的有效方法，但是当次数较高的时候，运算量太大，难以应用到实时图象处理系统，为此在分析多项式坐标变换算法的基础上，提出了一次多项式非均匀分片逼近算法。该算法首先将图象非均匀划分成矩形区域，在每个矩形区域内部用一个一次多项式逼近高次多项式。基于对图象畸变的主要因素即径向畸变的分析，该算法的图象划分规则能在保证逼近精度的前提下占用最少的保存模型参数的空间。该算法大大降低了运算量，将运算时间减少了近2／3，同时能很好地保证逼近精度，空间代价也限制在很小的范围内，试验结果表明该算法是图象几何修正的有效方法，具有良好的工程应用价值。     

基于子带可控响应功率的多声源定位方法

为了提高多个说话人情况下麦克风阵列的定位性能,提出基于子带可控响应功率的多声源定位算法。该算法将语音信号频域分为7个子带,在每个子带计算相位变换加权的可控响应功率函数,在声源空间搜索其最大值得到声源位置的初始估计。根据语音信号频率的稀疏性,这些初始估计包含多个声源的位置,运用会聚聚类算法得到最终的声源位置估计。仿真和实验表明,在有2个说话人,10 dB信噪比,较强混响的条件下,该算法比传统算法的定位正确率提高了约4%,额外率降低了约7%。     

结合非线性频谱与核主元分析的复杂系统故障诊断方法

传统非线性频谱分析方法对复杂系统进行故障诊断时,求解出的非线性频谱数据量庞大,不便于直接用于故障检测与分类识别.本文提出了一种非线性频谱特征与核主元分析(KPCA)结合的故障诊断方法,首先通过最小二乘算法估计出前3阶Volterra时域核,由多维傅立叶变换求取出广义频率响应函数,然后利用KPCA方法对谱数据进行压缩与提取谱特征,最后利用多分类最小二乘支持向量机进行多故障检测与识别.考虑到频谱数据具有非线性的特点,KPCA中的核函数选用由多项式函数与径向基函数构成的混合核函数,兼顾了局部特性与全局特性.论文基于非线性频谱数据,给出了核主元模型建立与在线故障诊断的具体算法.对非线性模拟电路和数控机床伺服传动系统进行了仿真实验,结果表明本文方法能够大幅度降低频谱数据维数,故障识别率高,是一种实用的故障诊断方法.     

一种新的三维模型水印频域算法

从提高三维模型数字水印的鲁棒性和效率出发．提出一种新的频域水印算法。该算法利用网格拓扑结构．构建出目标顶点所处的局部几何空间．进而提取出适用于DCT变换的信息序列,修改频域系数嵌入水印：然后通过逆向DCT转回空域信号,修改目标顶点在局部几何空间中的位置得到含水印模型。算法实现简单．水印检测无需预处理．且能有效抵抗平移、旋转、缩放、剪切、网格简化等攻击。实验结果表明,该算法具有很好的鲁棒性、水印的不可感知性、水印检测高效准确的优势。     

一种新的三维模型水印频域算法

从提高三维模型数字水印的鲁棒性和效率出发,提出一种新的频域水印算法。该算法利用网格拓扑结构,构建出目标顶点所处的局部几何空间,进而提取出适用于DCT变换的信息序列,修改频域系数嵌入水印;然后通过逆向DCT转回空域信号,修改目标顶点在局部几何空间中的位置得到含水印模型。算法实现简单,水印检测无需预处理,且能有效抵抗平移、旋转、缩放、剪切、网格简化等攻击。实验结果表明,该算法具有很好的鲁棒性、水印的不可感知性、水印检测高效准确的优势。