基于子编码和全编码联合惩罚的稀疏表示字典学习方法

针对目前稀疏表示字典学习的惩罚函数版本不一且各有优势的问题,提出基于子编码和全编码联合惩罚的稀疏表示字典学习方法,该方法在字典学习的目标函数中同时加入子编码惩罚函数和全编码惩罚函数。子编码惩罚函数使得学习后的字典在稀疏表示识别时可以用子字典的重构误差和子字典上编码系数的大小来识别,全编码惩罚函数则能直接利用整个字典上的编码系数来识别,通过联合这两个惩罚函数可以获得非常好的识别效果。为了验证所提方法的有效性,在语音情感库和人脸库上与最新的基于字典学习的稀疏表示识别方法 DKSVD和FDDL进行对比,并与著名的识别方法SVM和SRC进行比较,实验结果显示所提方法具有更好的识别性能。     

基于字典学习的软件缺陷检测算法

针对目前存在的字典学习方法不能有效构造具有鉴别能力字典的问题,提出具有鉴别表示能力的字典学习算法,并将其应用于软件缺陷检测。首先,重新构建稀疏表示模型,通过在目标函数中设计字典鉴别项学习具有鉴别表示能力的字典,使某一类的字典对于本类的样本具有较强的表示能力,对于异类样本的表示效果则很差;其次,添加Fisher准则系数鉴别项,使得不同类的表示系数具有较好的鉴别能力;最后对设计的字典学习模型进行优化求解,以获得具有强鉴别和稀疏表示能力的结构化字典。选择经过预处理的NASA软件缺陷数据集作为实验数据,与主成分分析（PCA）、逻辑回归、决策树、支持向量机（SVM）和代表性的字典学习方法进行对比,结果表明所提出的具有鉴别表示能力的字典学习算法的准确率与F-measure值均有提高,能在改善分类器性能的基础上提高检测精度。     

结构稀疏表示分类目标跟踪算法

为提高目标跟踪算法在复杂条件下的鲁棒性和准确性,研究了一种基于贝叶斯分类的结构稀疏表示目标跟踪算法。首先通过首帧图像获得含有目标与背景模板的稀疏字典和正负样本;然后采用结构稀疏表示的思想对样本进行线性重构,获得其稀疏系数;进而设计一款贝叶斯分类器,分类器通过正负样本的稀疏系数进行训练,并对每个候选目标进行分类,获得其相似度信息;最后采用稀疏表示与增量学习结合的方法对稀疏字典进行更新。将该算法与其他4种先进算法在6组测试视频中进行比较,实验证明了该算法具有更好的性能。     

面向乳腺病理图像分类的非相干字典学习及稀疏表示算法

针对乳腺病理图像分类,提出一种非相干字典学习及其稀疏表示算法.首先针对不同类别的图像,基于在线字典学习算法分别学习各类特定的子字典;其次利用紧框架建立一种非相干字典学习模型,通过交替投影优化字典的相干性、秩与紧框架性,从而有效地约束字典的格拉姆矩阵与参考格拉姆矩阵的距离,获得判别性更强的非相干字典;最后采用子空间旋转方法优化非相干字典的稀疏表示性能.利用乳腺癌数据集BreaKHis进行实验的结果证明,该算法所学习的非相干字典能平衡字典的判别性与稀疏表示性能,在良性肿瘤与恶性肿瘤图像分类上获得了86.0%的分类精度;在良性肿瘤图像中的腺病与纤维腺瘤的分类上获得92.5%的分类精度.     

Fisher准则下面向判别性特征的字典学习方法及其组织病理图像分类研究

针对当前面向组织病理图像特征提取的字典学习方法中存在着学习的无病字典与有病字典相似程度高,判别性弱的问题,本文提出一种新的面向判别性特征字典学习方法（Discriminative feature-oriented dictionary learning based on Fisher criterion,FCDFDL）.该方法基于Fisher准则构造目标函数的惩罚项,最小化学习字典的类内距离与最大化学习字典的类间距离,大大降低无病字典与有病字典间的相似性.同时,优化学习字典对同类样本的重构性能,并约束学习字典对非同类样本的重构性能.然后,利用本文学习的无病与有病字典对测试样本进行稀疏表示,采用重构误差向量的统计量构造分类器.最后,分别在ADL数据集与BreaKHis数据集上验证了本文方法的有效性.实验结果表明,本文学习字典的判别性更强,获得了更优的分类性能.     

基于鉴别性低秩表示及字典学习的鲁棒人脸识别算法

针对人脸识别中的图像存在噪声等情况,提出基于鉴别性低秩表示及字典学习的算法。使用鉴别性低秩子空间恢复算法(discriminative low-rank representation, DLRR)获得类别间尽可能独立且干净的训练样本,然后通过引入基于Fisher准则的字典学习(Fisher Discrimination Dictionary Learning, FDDL)方法得到结构化字典,其子字典对对应的类有较好的表示能力,约束编码系数具有较小类内散列度和较大类间散列度。最后对测试样本稀疏线性表示时正确类别的样本贡献更大。在标准人脸数据库上的实验结果表明该算法有较好性能。     

判别稀疏表示与在线字典学习的运动目标跟踪

针对传统稀疏表示不能有效区分目标和背景的缺点,提出一种判别稀疏表示算法,这种算法在传统稀疏表示目标函数中加入一个判别函数,大大降低干扰因素对目标跟踪的影响。基于判别稀疏表示和[?1]约束,提出一种在线字典学习算法升级目标模板,有效降低背景信息对目标模板的影响。提取目标梯度方向的直方图（HOG）特征,利用其对光照和形变等复杂环境具有较强鲁棒性的优点,实现对目标更稳定的跟踪。实验结果表明,与现有跟踪方法相比,该算法的跟踪效果更好。     

人脸识别中基于系数相似性的字典学习算法

在现有的基于稀疏表示分类算法的人脸识别中,使用通过稀疏学习得到的精简字典可以提高识别速度和精确度。metaface学习(Metaface Learning,MFL)算法在字典学习过程中没有考虑同类样本稀疏编码系数之间具有相似性的特点。为了利用这一信息来提高字典的区分性,提出了一种基于系数相似性的metaface学习(Coefficient-Simi-larity-based Metaface earning,CS-MFL)算法。CS-MFL算法的学习过程中,在更新稀疏表示系数阶段加入同类训练样本稀疏编码系数相似的约束项。为了求解包含系数相似性约束的新的最优化问题,将目标函数中的两个l2范数约束项进行合并,将原问题转化为典型l₂- l1问题进行求解。在不同的人脸库上进行实验,结果表明,提出的CS-MFL算法能够获得比MFL算法更高的识别率,说明由CS-MFL算法学习得到的字典更高效且更具区分性。     

带边界条件约束的非相干字典学习方法及其稀疏表示

从字典的相干性边界条件出发, 提出一种基于极分解的非相干字典学习方法(Polar decomposition based incoherent dictionary learning, PDIDL), 该方法将字典以Frobenius范数逼近由矩阵极分解获取的紧框架, 同时采用最小化所有原子对的内积平方和作为约束, 以降低字典的相干性, 并保持更新前后字典结构的整体相似特性. 采用最速梯度下降法和子空间旋转实现非相干字典的学习和优化. 最后将该方法应用于合成数据与实际语音数据的稀疏表示. 实验结果表明, 本文方法学习的字典能逼近等角紧框架(Equiangular tight-frame, ETF), 实现最大化稀疏编码, 在降低字典相干性的同时具有较低的稀疏表示误差.     

面向人脸识别的判别低秩字典学习算法

人脸识别是计算机视觉和模式识别领域的一个研究热点,有着十分广泛的应用前景.人脸识别任务在训练样本和测试样本同时包含噪声的情况下存在识别精度不高的问题,为此本文提出一个新的判别低秩字典学习和低秩稀疏表示算法（Discriminative Low-Rank Dictionary Learning for Low-Rank Sparse Representation,DLRD_LRSR）.本文方法在模型中约束每个子字典和稀疏表示低秩避免噪声干扰,并引入了判别重构误差项增强系数的判别性.为验证算法的有效性,本文在3个公开人脸数据集上进行了实验评估,结果表明与现有字典学习算法相比,本文算法能够更好的解决训练样本和测试样本同时存在噪声的人脸识别问题.     

基于层次结构化字典学习的人脸表情识别*

针对传统稀疏表示方法构建的字典不具备判别性的问题,以K-SVD算法为基础,对判别字典的构建和分类求解进行了研究,提出一种基于层次结构化字典学习的表情识别方法。先将训练样本切割出眼眉、脸颊和嘴三部分,对分割的各部分利用K-SVD算法得到块字典向量,再用层次分析法的权重赋值方法求块字典向量的权重值,构成各类子字典。将所有的子字典进行联合,用结构化字典学习算法求解。测试样本的归类取决于求解结果重构的效果。在JAFFE和CK表情库上的实验表明,该算法在保证了字典判别性的同时,也达到了较高的识别率。     

PolSAR ship detection based on the contrast enhancement of polarimetric scattering differences

The ship detection in polarimetric synthetic aperture radar (PolSAR) mode is a hot topic in recent years, because of the diversity of polarimetric scattering mechanisms between ship targets and sea clutter. To improve the detection performance of ship targets, this paper mainly develops the ship detection method based on the contrast enhancement utilizing the polarimetric scattering difference. The algorithm first enhances the target signal utilizing the scattering difference of the polarimetric coherency matrix between ship targets and sea clutter, and then a simple threshold is applied to distinguish the ship targets from the sea clutter. Finally, real PolSAR datasets recorded by AirSAR system are used to evaluate the effectiveness of the proposed detection method. Compared with other detection methods, experimental results indicate that the proposed method can effectively improve the detection performance of ship targets.     

Source cell phone verification from speech recordings using sparse representation

Source recording device recognition is an important emerging research field in digital media forensics. The literature has mainly focused on the source recording device identification problem, whereas few studies have focused on the source recording device verification problem. Sparse representation based classification methods have shown promise for many applications. This paper proposes a source cell phone verification scheme based on sparse representation. It can be further divided into three schemes which utilize exemplar dictionary, unsupervised learned dictionary and supervised learned dictionary respectively. Specifically, the discriminative dictionary learned by supervised learning algorithm, which considers the representational and discriminative power simultaneously compared to the unsupervised learning algorithm, is utilized to further improve the performances of verification systems based on sparse representation. Gaussian supervectors (GSVs) based on MFCCs, which have shown to be effective in capturing the intrinsic characteristics of recording devices, are utilized for constructing and learning dictionary. SCUTPHONE, which is a corpus of speech recordings from 15 cell phones, is presented. Evaluation experiments are conducted on three corpora of speech recordings from cell phones and demonstrate the effectiveness of the proposed methods for cell phone verification. In addition, the influences of number of target examples in the exemplar dictionary and size of the unsupervised learned dictionary on source cell phone verification performance are also analyzed.     

Single-channel blind source separation based on joint dictionary with common sub-dictionary

The cross projection engenders when mixed speech signal is represented over joint dictionary because of the bad distinguishing ability of joint dictionary in single-channel blind source separation (SBSS) using sparse representation theory, which leads to bad separation performance. A new algorithm of constructing joint dictionary with common sub-dictionary is put forward in this paper to this problem. The new dictionary can effectively avoid being projected over another sub-dictionary when a source signal is represented over joint dictionary. In the new algorithm, firstly we learn identify sub-dictionaries using source speech signals corresponding to each speaker. And then we discard similar atoms between two identity sub-dictionaries and construct a common sub-dictionary using these similar atoms. Finally, we combine those three sub-dictionaries together into a joint dictionary. The Euclidean distance among two atoms is used to measure the correlation of them in different identity sub-dictionaries, and similar atoms are searched based on the correlation. In testing stage, each source can be reconstructed with the projection coefficients corresponding to individual sub-dictionary and the common sub-dictionary. Contrast experiments tested in speech database show that the algorithm proposed in this paper performs better, when the Signal-to-Noise Ratio (SNR) is used to measure separation effect. The algorithm set out in this paper has lower time complexity as well.     

Gabor特征集结合判别式字典学习的稀疏表示图像识别

稀疏编码中字典的选择无论对图像重建还是模式分类都有重要影响,为此提出Gabor特征集结合判别式字典学习的稀疏表示图像识别算法.考虑到Gabor局部特征对光照、表情和姿态等变化的鲁棒性,首先提取图像对应不同方向、不同尺度的多个Gabor特征;然后将降维的增广Gabor特征矩阵作为初始特征字典,通过对该字典的学习得到字典原子对应类别标签的新结构化字典,新字典中特定类的子字典对相关的类具有好的表示能力,同时应用Fisher判别约束编码系数,使它们具有小的类内散度和大的类间散度;最后同时用具有判别性的重构误差和编码系数来进行模式分类.基于3个数据库的实验结果表明本文方法具有可行性和有效性.     

图正则化字典对学习的轻度认知功能障碍预测

针对字典对学习（DPL）方法只考虑了同类子字典的重构误差和不同类表示系数的稀疏性,没有考虑图像间的几何近邻拓扑关系的问题。通过近邻保持使得在同类近邻投影系数之间的距离较小,而不同类投影系数之间的距离大,能够有效提高字典对学习算法的分类性能,基于此提出了基于几何近邻拓扑关系的图正则化的字典对学习（GDPL）算法。在ADNI1数据集上对轻度认知功能障碍预测的实验表明,使用GDPL算法学习的编码系数作为特征预测的准确率（ACC）和ROC曲线下的面积（AUC）比使用结合生物标志作为特征预测的准确率提高了2%~6%,使用GDPL算法比DPL算法的实验结果也有提高。     

Sparse Representation Based Fisher Discrimination Dictionary Learning for Image Classification

The employed dictionary plays an important role in sparse representation or sparse coding based image reconstruction and classification, while learning dictionaries from the training data has led to state-of-the-art results in image classification tasks. However, many dictionary learning models exploit only the discriminative information in either the representation coefficients or the representation residual, which limits their performance. In this paper we present a novel dictionary learning method based on the Fisher discrimination criterion. A structured dictionary, whose atoms have correspondences to the subject class labels, is learned, with which not only the representation residual can be used to distinguish different classes, but also the representation coefficients have small within-class scatter and big between-class scatter. The classification scheme associated with the proposed Fisher discrimination dictionary learning (FDDL) model is consequently presented by exploiting the discriminative information in both the representation residual and the representation coefficients. The proposed FDDL model is extensively evaluated on various image datasets, and it shows superior performance to many state-of-the-art dictionary learning methods in a variety of classification tasks.     

HEp-2 cell pattern classification with discriminative dictionary learning

The paper presents a supervised discriminative dictionary learning algorithm specially designed for classifying HEp-2 cell patterns. The proposed algorithm is an extension of the popular K-SVD algorithm: at the training phase, it takes into account the discriminative power of the dictionary atoms and reduces their intra-class reconstruction error during each update. Meanwhile, their inter-class reconstruction effect is also considered. Compared to the existing extension of K-SVD, the proposed algorithm is more robust to parameters and has better discriminative power for classifying HEp-2 cell patterns. Quantitative evaluation shows that the proposed algorithm outperforms general object classification algorithms significantly on standard HEp-2 cell patterns classifying benchmark¹ and also achieves competitive performance on standard natural image classification benchmark.