Image annotation using multi-view non-negative matrix factorization with different number of basis vectors

Automatic Image Annotation (AIA) helps image retrieval systems by predicting tags for images. In this paper, we propose an AIA system using Non-negative Matrix Factorization (NMF) framework. The NMF framework discovers a latent space, by factorizing data into a set of non-negative basis and coefficients. To model the images, multiple features are extracted, each one represents images from a specific view. We use multi-view graph regularization NMF and allow NMF to choose a different number of basis vectors for each view. For tag prediction, each test image is mapped onto the multiple latent spaces. The distances of images in these spaces are used to form a unified distance matrix. The weights of distances are learned automatically. Then a search-based method is used to predict tags based on tags of nearest neighbors’. We evaluate our method on three datasets and show that it is competitive with the current state-of-the-art methods.     

基于非负矩阵分解的图像脆弱水印算法

提出一种基于非负矩阵分解NMF(Non-negative Matrix Factorization)的脆弱数字水印算法。算法利用用户密钥构造NMF基矩阵,并在图像NMF分解过程中保持不变,二值水印图像嵌入NMF分解系数矩阵。实验结果本算法具有较强的鲁棒性,同时用户密钥保证的算法的脆弱性。     

以非负矩阵分解提取局部特征的SAR目标稀疏表示分类

合成孔径雷达( SAR)目标分类是自动目标识别系统的核心功能之一,对于战场监视等应用具有重要意义。利用SAR图像局部散射明显的特点,提出了通过训练样本的非负矩阵分解获得低维数局部特征编码,并以该编码作为字典进行稀疏表示分类的方法。采用Gotcha项目民用车辆目标的实测数据进行了验证,结果显示在不同信噪比条件下该方法的分类正确率均优于广泛采用的由降采样、随机投影、主成分分析提取低维数特征的稀疏表示分类方法,表明了该方法的性能优势。另外,还通过实验对比分析了非负约束的稀疏表示与标准稀疏表示在分类性能上的差别,结果显示非负约束的稀疏表示导致分类正确率下降,故针对分类问题不宜在稀疏表示时进行非负约束。     

A SCBSS methodology for time-frequency overlapped signals using non-negative matrix factorisation

This article presents a new methodology for single-channel blind signal separation (SCBSS) of time-frequency overlapped signals in electromagnetic surveillance domain. The proposed algorithm is based on non-negative matrix factorisation (NMF) through a dynamical embedding framework. This method is deterministic requiring no priori information of the original signals. Firstly, a series of delay vectors of the single-channel recording are selected to construct an appropriate dynamical embedding matrix. NMF algorithm is used to decompose the dynamical embedding matrix. Then, a convenient expansion basis of original sources can be obtained, which will be projected back to the original measurement space according to the dynamical embedding framework. The power density spectrum of the projected components and the independent nature of original signals are employed to extract the interested signals from measurement space. The feasibility and effectiveness of proposed algorithm are proved by computer simulation.     

基于非负矩阵分解算法进行盲信号分离

独立分量分析(ICA)已被广泛运用于线性混合模型的盲源分离问题,但却有两个重要的限制:信源统计独立和信源非高斯分布。然而更有意义的线性混合模型是:观测信号是非负信源的非负线性混合,信源之间可以统计相关且可以为高斯分布。本文针对盲源分离问题,提出了一种运用新近国际上提出的一种非负矩阵分解算法(NMF算法)进行统计相关信源的盲源分离方法,该方法没有信源统计独立和信源非高斯分布的限制,只要信源之间没有一阶原点统计相关,则可很好实现对信源的分离。大量仿真及与传统ICA进行盲源分离的比较,验证了运用NMF进行包括统计相关信源和高斯分布信源的盲源分离的可行性和有效性。     

Topology preserving non-negative matrix factorization for face recognition.

In this paper, a novel topology preserving non-negative matrix factorization (TPNMF) method is proposed for face recognition. We derive the TPNMF model from original NMF algorithm by preserving local topology structure. The TPNMF is based on minimizing the constraint gradient distance in the high-dimensional space. Compared with L(2) distance, the gradient distance is able to reveal latent manifold structure of face patterns. By using TPNMF decomposition, the high-dimensional face space is transformed into a local topology preserving subspace for face recognition. In comparison with PCA, LDA, and original NMF, which search only the Euclidean structure of face space, the proposed TPNMF finds an embedding that preserves local topology information, such as edges and texture. Theoretical analysis and derivation given also validate the property of TPNMF. Experimental results on three different databases, containing more than 12,000 face images under varying in lighting, facial expression, and pose, show that the proposed TPNMF approach provides a better representation of face patterns and achieves higher recognition rates than NMF.     

基于非负矩阵分解的半监督动态社团检测

如何有效融合不同时刻的网络结构信息,是影响复杂网络中动态社团检测算法检测性能的关键和难点。基于此,提出了一种基于非负矩阵分解的半监督动态社团检测方法SDCD-NMF,该方法首先有效提取了历史时刻所包含的稳定结构单元,然后将其作为正则化监督项,指导当前时刻的网络社团检测。在真实网络数据集上的实验表明,所提方法与已有方法相比具备更高的社团划分质量,更有利于探索网络的演变与发展规律。     

基于非负矩阵分解的社区检测算法

为了了解复杂网络的特性,研究了复杂网络中的社区交叠现象,将非负矩阵分解算法用于社区检测问题。而传统的用于社区检测SNMF模型是通过离散化参数的取值范围,然后遍历得到参数的最优值,对参数的优化方法不能准确而快速搜索到最优解。利用遗传算法对参数进行优化,能够准确地找到参数的最优解,从而得到最优的社区划分。并且能够检测出交叠节点和异常节点,该算法也适应于大规模的数据。     

改进的非负矩阵分解在基因表达数据中的应用

对非负矩阵分解算法的发展历史做了简要综述,并介绍了非负矩阵分解算法的基本原理,及一种改进的非负矩阵分解算法,并将此方法应用于一组阿尔茨海默症微阵列数据的聚类中,将此结果与传统的非负矩阵分解算法进行了比较,证明了算法的有效性.     

非对称代价函数的稀疏卷积非负矩阵分解方法

提出一种基于非对称代价函数的稀疏卷积非负矩阵分解方法。该方法利用板仓-斋藤距离作为目标代价函数来衡量目标矩阵与重建矩阵的差异,使得较小的矩阵元素具有较小的重建误差,并且该代价函数具有尺度不变性的特点。为了考察其在弱语音成分重建方面的优势,将本文提出的算法应用于耳语音谱分解及重建实验。实验结果表明,与基于欧氏距离和基于Kullback-Leibler(K-L)散度的卷积非负矩阵分解算法相比,本文算法对于弱语音成分具有更好的重构效果,重建后的语音信号具有较大的可懂度。      

基于稀疏约束非负矩阵分解的人脸特征提取算法

非负矩阵分解(NMF)是最近流行的一种提取数据局部特征的算法,虽该算法已成功用于多种领域,但其并不能总是最好地表示局部特征。针对上述问题,文中在非负矩阵分解的同时加入稀疏的限制,并通过限制稀疏度从而提高局部特征的提取效果。通过在人脸图片上的实验可明显看出,加入稀疏限制的非负矩阵分解能更清楚地提取出所需的局部特征,以便于后续针对特征进行的各种工作。     

基于局部Walsh变换和非负矩阵分解的脑白质图像分割

脑白质病变诊断是医学研究和病理分析的重要方面。颅脑核磁共振图像的白质分割在诊断中起着非常重要的作用,其分割的准确性直接影响后续的分析和诊断研究。本文提出了一种基于局部Walsh变换和非负矩阵分解的大脑核磁共振图像白质分割算法。算法首先对颅脑图像进行局部Walsh变换,选择鉴别性能好的特征得到特征矩阵,然后对其进行非负矩阵分解并得到白质的分割结果。实验表明,本方法计算简单,精度比较高,可以得到比较理想的分割结果。     

Blind nonlinear hyperspectral unmixing based on constrained kernel nonnegative matrix factorization

Spectral unmixing has been a useful technique for hyperspectral data exploration since the earliest days of imaging spectroscopy. As nonlinear mixing phenomena are often observed in hyperspectral imagery, linear unmixing methods are often unable to unmix the nonlinear mixtures appropriately. In this paper, we propose a novel blind unmixing algorithm, constrained kernel nonnegative matrix factorization, which obtains the endmembers and corresponding abundances under nonlinear mixing assumptions. The proposed method exploits the nonlinear structure of the original data through kernel-induced nonlinear mappings and one need not know the nonlinear model. In order to improve its performance further, two auxiliary constraints, namely simplex volume constraint and abundance smoothness constraint, are also introduced into the algorithm. Experiments based on synthetic datasets and real hyperspectral images were performed to evaluate the validity of the proposed method.     

基于核主元分析法的组合核函数改进算法

提出了一种新型的组合核函数应用于构建支持向量机当中.这种组合核函数将高斯核函数与多项式核函数各自的特点融合在一起,构建了一种兼具内推和外推性能的核函数.经实验验证,将这一核函数应用在核主元分析法中,可以有效地提高识别精确度和效率.     

Sensor network localization using kernel spectral regression

This paper addresses the localization problem in wireless sensor networks using signal strength. We use a kernel function to measure the similarities between sensor nodes. The kernel matrix can be naturally defined in terms of the signal strength matrix. We show that the relative locations of sensor nodes can be obtained by solving a dimension reduction problem. To capture the structure of the whole network, we use the kernel spectral regression (KSR) method to estimate the relative locations of the sensor nodes. Given sufficient anchor nodes, the relative locations can be aligned to global locations. The key benefits of adopting KSR are that it allows us to define a graph to optimally preserve the topological structure of the sensor network, and a kernel function can capture the nonlinear relationship in the signal space. Simulation results show that we can achieve small average location error with a small number of anchors. We also compare our method with several related methods, and the results show that KSR is more efficient than the others in our simulated sensor networks. Copyright © 2009 John Wiley & Sons, Ltd.     

基于单核和组合核函数在垃圾邮件过滤中的比较应用

大多数文本为高维且线性不可分。针对中文邮件,首先阐述了邮件预处理的相关方法,利用TF－TDF将邮件向量化。分析了多种常用核函数在SVM中应用于垃圾邮件过滤。阐述了全局核函数和局部核函数的特点,主要针对全局核函数－多项式（Poly）核函数和局部核函数－径向基核（RBF）函数在垃圾邮件分类的准确性做了比较,综合分析后组合两种核函数。实验证明,组合核函数在性能上优于单个核函数,具有较好的学习能力和泛化能力。     

基于采样协方差矩阵的混合核SVM高效频谱感知

近年来随着盲检测算法的提出,越来越多的基于采样协方差矩阵的盲检测算法应用于频谱感知。针对其检测门限是近似值,检测性能会受到影响等问题,提出了基于采样协方差矩阵的混合核函数的支持向量机（support vector machine,SVM）高效频谱感知,通过感知信号采样协方差矩阵的最大最小特征值（maximum minimum eigenvalue,MME）和协方差绝对值（covariance absolute value,CAV）提取的统计量作为SVM的特征向量并训练其生成频谱感知的分类器,无需计算检测门限并且特征提取减少了样本集的大小。利用遗传算法（genetic algorithm,GA）优化混合核函数的SVM的参数。实验结果表明,该方法比MME算法和CAV算法的检测概率有所提高,并且比SVM减少了感知时间,具有良好的实用性。     

High-quality voice conversion system based on GMM statistical parameters and RBF neural network

A voice conversion （VC） system was designed based on Gaussian mixture model （GMM） and radial basis function （RBF） neural network. As a voice conversion model, RBF network needs quantities of training data to improve its performance. For one speech, the networks trained by different segments of data have different transformation effects. Since trying segment by segment to obtain the best conversion effect is complex, a conversion method was proposed, that uses GMM for statistics before training RBF network to aim at the problem. The speech transformation and representation using adaptive interpolation of weighted spectrum （STRAIGHT） model is used for accurate extraction of vocal tract spectrum. Then GMM is used to classify the numerous spectral parameters. The obtained mean parameters were trained in RBF network. Experiment reveals that, the soft classification ability of GMM can promptly realize the reduction and classification of training data under the premise of ensuring the training effect. The selection complexity is decreased thereafter. Compared to the conventional RBF network training methods, this method can make the transformation of spectral parameters more effective and improve the quality of converted speech.