Partial Oblique Projection Learning for Optimal Generalization

In some viewpoints, supervised learning is discussed in the framework of function approximation, which means that different criteria result in learning methods of different abilities in generalization [1]. From the standpoint of the original space to which the desired function belongs, projection-based criterion aims directly at the generalization ability[2]. The projection concept is applied to projection learning (PL), partial projection learning (PTPL), and averaged projection learning (A…     

Incremental-based Nonlinear Detection Algorithm for MIMO System

A novel incremental nonlinear detection algorithm is presented for Multiple-Input Multiple-Output （MIMO） system. In this algorithm, the received data at multiple receiver antennas are nonlinearly mapped and then summed with weights. The weight coefficients are incrementally computed to avoid direct computation of the inverse of a matrix, which greatly reduce the computational complexity. Simulation and comparison show that the proposed algorithm can obtain better performance of Bit Error Rate （BER） than linear Minimum Mean Square Error （MMSE）.     

基于核方法的雷达目标一维距离像识别

由于雷达目标及其所处环境的复杂性,导致不同目标之间的关系往往是非线性的.研究基于核的非线性方法,并将其应用于雷达目标一维距离像识别.核Fisher判别分析(KFDA)是一种抽取非线性特征的最有效方法之一,但它往往会面临小样本问题.针对此问题,给出一种null-KFDA方法,对距离像进行特征提取.然后,采用一种新的核非线性分类器——KNR(kernel-based nonlinear representor),对所提取的特征进行分类.对3种飞机的实测距离像进行实验,结果验证了null-KFDA的有效性.此外,与非线性支持向量机(SVM)和径向基函数神经网络(RBFNN)相比,KNR分类器具有更优的识别性能.     

一种利用相位信息的雷达目标成像识别方法

高距离分辨雷达（ＨＲＲ）的回波中蕴含丰富的目标信息,充分利用目标回波并采用恰当的特征抽取方法,可以有交地识别目标。本文提出一种利用相位信息和正则变换的目标识别方法。该方法基于雷达目标距离剖面像的幅度和相位矢量（幅相矢量）,首先由各训练目标在不同方位瓷态角时的幅相衡量构成综合矩阵并对之作正则变换建立正则子空间,然后奖每类训练目标各方位姿态的幅相矢量向该子空间投影形成子像,取其平均结果作为该目标的库特     

一种多用户VBLAST MIMO系统快速非线性检测算法

提出了一种针对多用户CDMA VBLAST MIMO系统结构的快速非线性检测算法,该算法首先通过核方法将接收端多个天线收到的信号进行非线性映射,得到新的数据空间,然后对其进行加权求和得到输出信号,最后形成系统判决输出。为了避免直接矩阵求逆运算,权系数的求解采用递增方式进行,这将显著减小算法计算量和复杂性。仿真结果表明:提出的算法在误码率方面能够取得明显优于常规线性MMSE算法的性能,算法的有效性得到验证。     

基于多帧融合及四参数仿射模型的图像超分辨

运动参数估计和复原是多帧图像超分辨重构中最重要的两个环节,其中经典的Fourier-Mellin变换方法于频域采用对数极坐标形式和相位相关方法结合来估计运动参数。相位相关是整像素级平移参数估计方法,将其改进为亚像素级平移参数估计方法,以提高旋转、缩放参数的估计精度。对于复原算法,在讨论基于局部信息的传统双三次插值超分辨重构方法的基础上,重点探讨基于全局信息的Kriging插值超分辨重构和核非线性回归（KNR）超分辨重构方法。实验结果表明,探讨的参数估计方法和超分辨重构方法是有效的。     

SAR图像去噪的非局部均值算法改进

针对非局部均值算法在合成孔径雷达(Synthetic Aperture Radar,SAR)图像的边缘和纹理区域去噪效果较模糊的问题,探讨改进了非局部均值算法的权重部分,把高斯加权的空间距离与非局部均值算法的权重系数相结合,构造新的权重系数,提出了一种基于改进非局部均值算法的SAR图像去噪算法.首先,使用对数变换处理图...     

子空间法雷达目标一维像识别研究

基于高分辨一维像,研究特征子空间法和正则子空间法在雷达目标识别中的应用。针对一维像敏感于目标姿态变化的特点,提出一种子空间串识别法,将所有姿态范围划分为一定数量的模区,在每模区建立各类目标的子空间.对未知目标,所处模区由雷达测定后,其一维像映射到该模区各类目标的特征子空间进行识别分类一单模区搜索准则.模拟和实测数据实验表明所提出方法是有效的.     

Incremental POP Learning

In neural network learning, it is practically necessary to retrain a trained network when novel data become available. Retraining from the scratch is highly inefficient in memory and computation while adjusting adaptively the trained result by the new data is a natural avenue and it is called incremental learning[1]. Recently, incremental learning becomes an interest issue in neural networks, and also many methods have been devised[1~3]. These methods are divided into two categories according …     

A Kernel-Based Nonlinear Representor with Application to Eigenface Classification

Classifiers play a paramount role in patternrecognition. In some of the popular methods, such asthe Parzen classifier and a nonlinear support vectorclassifier[1,2], the solutions are expressed as a nonlinearfunction f expanded by a kernel function, k, of thespace to which f belongs, 1( ) ( , )Mj j jf a k b=x = ∑x x ,where xj is the j-th sample point with j=1,2,…, M, b isa constant which can be set to zero in someapplications. The set of coefficients, Mjja1{}=, isdecided by the nature of th…