Mahdi Abavisani, Vishal M. Patel · IEEE Signal Processing Letters · 2019
We present a transductive deep learning-based formulation for the sparse representation-based classification (SRC) method. The proposed network consists of a convolutional autoencoder along with a fully connected layer. The role of the autoencoder network is to learn robust deep features for classification. On the other hand, the fully connected layer, which is placed in between the encoder and the decoder networks, is responsible for finding the sparse representation. The estimated sparse codes are then used for classification. Various experiments on three different datasets show that the proposed network leads to sparse representations that give better classification results than state-of-the-art SRC methods. The source code is available at: github.com/mahdiabavisani/DSRC.
Cited by 35Open access
Ke Huang, Selin Aviyente · The MIT Press eBooks · 2007
In this paper, application of sparse representation (factorization) of signals over an overcomplete basis (dictionary) for signal classification is discussed.Searching for the sparse representation of a signal over an overcomplete dictionary is achieved by optimizing an objective function that includes two terms: one that measures the signal reconstruction error and another that measures the sparsity.This objective function works well in applications where signals need to be reconstructed, like coding and denoising.On the other hand, discriminative methods, such as linear discriminative analysis (LDA), are better suited for classification tasks.However, discriminative methods are usually sensitive to corruption in signals due to lacking crucial properties for signal reconstruction.In this paper, we present a theoretical framework for signal classification with sparse representation.The approach combines the discrimination power of the discriminative methods with the reconstruction property and the sparsity of the sparse representation that enables one to deal with signal corruptions: noise, missing data and outliers.The proposed approach is therefore capable of robust classification with a sparse representation of signals.The theoretical results are demonstrated with signal classification tasks, showing that the proposed approach outperforms the standard discriminative methods and the standard sparse representation in the case of corrupted signals.
Cited by 614
Long H. Ngo, Marie Luong, Nikolay Sirakov et al. · 2018
This paper improves the conventional sparse representation based classification (SRC) method, through incorporating wavelet coefficients. For this reason, the proposed method is called Sparse Representation Wavelet based Classification (SRWC). In the present study, we fuse the image features described by the complementary information from the low sub-band of the wavelet coefficients and sparse representation to outperform the conventional SRC according to accuracy. This holds because the wavelets promote sparsity and provide structural information about the image, which increases the accuracy of classification. To validate the capabilities and underline the advantages of the novel SRWC, we conducted an extensive number of experiments using publicly available datasets and compared our results with contemporary methods.
Cited by 10
Krishan Sharma, Renu Rameshan · 2017
Block-sparse representation based classifier (BSRC), an extension of sparse representation based classification (SRC), shows good performance for face recognition task with small amount of training data. However, BSRC does not handle the inter-block and intra-block non-linearities present in the input feature space. As most of the real world data is non-linear, this paper presents a non-linear extension of BSRC, kernel block-sparse representation based classifier (KBSRC), using the kernel trick. We transform the data non-linearity to a higher dimensional space. Two convex optimization problems based on block sparsity are solved in this kernel feature space. Dimensionality reduction technique in kernel feature space is also applied to avoid the increase in computational complexity of the algorithm. In the reduced kernel feature subspace, a test sample is assigned a label based on block-sparse coefficients. To validate the proposed approach, experimentation over Extended Yale B Face, ISOLET and MNIST datasets is performed and it shows significant improvement in classification performance with less training data than state-of-the-art methods.
Cited by 1
Keyang Cheng A, Xiaoyang Tan A, Communicated Tao Mei · 2014
Attribute reduce a b s t r a c t Classification for flowers is a very difficult task. Traditional methods need to built a classifier for each flower category, and obtain large number of flower samples to train these classifiers. In practice, many different types of flowers make the job become very difficult and boring. In this work, we present an attribute based approach for flowers recognition. Particularly, instead of training for a specific category of flowers directly based on manually designed features such as SIFT and HoG, we extract a series of visual attributes from a given set of flower images and generalize these to new images with possibly unknown flowers. A recently proposed sparse representations classification scheme is employed to predict the attributes of a given flower image from any category. In addition, we use a genetic algorithm to find the most discriminative attributes among others for better performance during the stage of flower classification. The effectiveness of the proposed method is validated on a publicly available flower classification database with promising results. & 2014 Elsevier B.V. All rights reserved. 1.
Cited by 0
Heng Zhang, Vishal M. Patel, Sumit Shekhar et al. · 2015
In recent years, sparse representation and dictionary learning methods have produced state-of-the-art results in many biometric recognition problems such as face, gait and iris recognition. However, when sparse representation-based classification methods are confronted with situations where the training data has different distribution than the test data, their performance degrades significantly. In this paper, we propose a general sparse representation-based classification method that learns projections of data in a space where the sparsity of data is maintained. We propose an efficient iterative procedure for solving the proposed optimization problem. One of the key features of the proposed method is that it is computationally efficient as the learning is done in the lower-dimensional space. Various experiments on mobile active authentication datasets consisting of face and screen touch gestures show that our method is able to capture the meaningful structure of data and can perform significantly better than many competitive domain adaptation algorithms.
Cited by 32
Ehsan Elhamifar, Renè Vidal · 2011
In many problems in computer vision, data in multiple classes lie in multiple low-dimensional subspaces of a high-dimensional ambient space. However, most of the existing classification methods do not explicitly take this structure into account. In this paper, we consider the problem of classification in the multi-sub space setting using sparse representation techniques. We exploit the fact that the dictionary of all the training data has a block structure where the training data in each class form few blocks of the dictionary. We cast the classification as a structured sparse recovery problem where our goal is to find a representation of a test example that uses the minimum number of blocks from the dictionary. We formulate this problem using two different classes of non-convex optimization programs. We propose convex relaxations for these two non-convex programs and study conditions under which the relaxations are equivalent to the original problems. In addition, we show that the proposed optimization programs can be modified properly to also deal with corrupted data. To evaluate the proposed algorithms, we consider the problem of automatic face recognition. We show that casting the face recognition problem as a structured sparse recovery problem can improve the results of the state-of-the-art face recognition algorithms, especially when we have relatively small number of training data for each class. In particular, we show that the new class of convex programs can improve the state-of-the-art face recognition results by 10% with only 25% of the training data. In addition, we show that the algorithms are robust to occlusion, corruption, and disguise.
Cited by 214
Hao Zheng, Fan Liu, Zhong Jin · Communications in computer and information science · 2012
Cited by 4
Soheil Bahrampour, Nasser M. Nasrabad, Asok Ray · WORLD SCIENTIFIC eBooks · 2016
Cited by 1
Hongwei Mo, Chaomin Luo, Gene Eu Jan · 2017
Classifying motor imagery brain signals where the signals are obtained based on imagined movement of the limbs is a major step in developing Brain Computer Interfaces (BCIs). Features from a small spatial region are approximated by a sparse linear combination of few atoms from a multi-class dictionary constructed from the features of the electroencephalography (EEG) training signals for each class. It is used to classify the signals based on the pattern of their sparse representation using a support vector classifier. The results show that the sparse representation classification algorithm is very effective for the pattern classification precision of the multi-channel EEG.
Cited by 4
Shaopeng Liu, Robert X. Gao, Dinesh John et al. · 2012
This paper presents a pattern classification method based on sparse representation. This new method bypasses the need for feature extraction and selection that are typically presented in the conventional classification methods, and performs classification using raw sensor signals directly. The performance of this new method is evaluated in the context of human physical activity assessment. Experimental results obtained from 105 human subjects demonstrate higher discriminative power than using the conventional k-nearest neighbor algorithm, verifying the effectiveness of the sparse representation method.© 2012 ASME
Cited by 0
Jin Huang · UTA ResearchCommons (University of Texas Arlington) · 2011
Image classification, such as face recognition and scene categorization, is an important research area in computer vision over the last decade. It has been successfully applied to many image analysis applications. Images usually have a large number of features, hence the dimensionality reduction methods are often employed before the subsequent classification to improve the classification results. A lot of methods have been proposed, including but not limited to PCA, ICA, LDA, and Bayesian Framework. Recently, compressive sensing and sparse learning have been widely studied and applied into computer vision research. Ma et al. suggested a new method called Sparse Representation Classification (SRC). This new framework provides new insights into two critical issues in image classification: feature extraction and robustness to occlusion. Motivated by this method, we proposed a new method called embedded sparse representation. This masterpiece combined the dimension reduction and classification into one. We proposed three possible objective functions and discussed the possible optimization ways to tackle them. During the optimization, we used Gibbs Optimization method to alternatively find the optimal subspace representation and the sparse weight factor. In the experiments, we verified the convergence of our method. Our method successfully extracted the subspace structure and got a better performance than SRC and other classical methods. It has also been shown that our method has the potential to be extended to other general high dimensional data. The possible improvement and future work have also been discussed.
Cited by 0Open access
Jian Lai · 2015
Building a computer as intelligent as or more intelligent than human is the ultimate goal of machine learning. To explore this powerful ability in computer vision, it inspires numerous interesting and useful research areas and applications. Face recognition (FR), as one of the most challenging problems in machine learning and computer vision, provides a good test bed to show how intelligent a machine can be. Moreover, it shows huge potential for commercial and surveillance, such as access control, computer-human interaction and anti-terrorist. FR remains a hot research topic though it has been studied for more than three decades. One important reason is that there is still large gap between human and machine to recognize the identity of a face image due to extreme illumination, expression and pose variation, disguise, and occlusion et al..
Cited by 0Open access
Chang Li, Yong Ma, Xiaoguang Mei et al. · IEEE Geoscience and Remote Sensing Letters · 2016
Recently, the sparse representation-based classification (SRC) methods have been successfully used for the classification of hyperspectral imagery, which relies on the underlying assumption that a hyperspectral pixel can be sparsely represented by a linear combination of a few training samples among the whole training dictionary. However, the SRC-based methods ignore the sparse representation residuals (i.e., outliers), which may make the SRC not robust for outliers in practice. To overcome this problem, we propose a robust SRC (RSRC) method which can handle outliers. Moreover, we extend the RSRC to the joint robust sparsity model named JRSRC, where pixels in a small neighborhood around the test pixel are simultaneously represented by linear combinations of a few training samples and outliers. The JRSRC can also deal with outliers in hyperspectral classification. Experiments on real hyperspectral images demonstrate that the proposed RSC and JRSRC have better performances than the orthogonal matching pursuit (OMP) and simultaneous OMP, respectively. Moreover, the JRSRC outperforms some other popular classifiers.
Cited by 100
Xiangping Sun, Jin Wang, Mary She et al. · Neurocomputing · 2013
Cited by 17
Ulises Rodríguez‐Domínguez, Oscar Dalmau · Signal Image and Video Processing · 2020
Cited by 3
Sanchez Arias, Reinaldo · DigitalCommons@UTEP (The University of Texas at El Paso) · 2011
In pattern recognition and machine learning, a classification problem refers to finding an algorithm for assigning a given input data into one of several categories. Many natural signals are sparse or compressible in the sense that they have short representations when expressed in a suitable basis. Motivated by the recent successful development of algorithms for sparse signal recovery, we apply the selective nature of sparse representation to perform classification. In order to find such sparse linear representation, we implement an l1-minimization algorithm. This methodology overcomes the lack of robustness with respect to outliers. In contrast to other classification algorithms such as Support Vector Machines (SVM), no model selection dependence is involved. The minimization algorithm is a convex relaxation-like algorithm that has been proven to efficiently recover sparse signals. To study its performance, the proposed method is applied to six tumor gene expression datasets with a large number of features but few samples. Our numerical results compare favorably with various SVM methods. We also test the effectiveness of our classification algorithm in the Fisher's Iris dataset where a large number of samples but a small number of features are available. Since the process and techniques for acquiring and analyzing data advance every day at high rates, we need to manage and analyze large amounts of data for several different scientific problems. Future work aims to study the performance of our classification method when dimensionality reduction techniques are applied, including feature selection and feature extraction strategies.
Cited by 0Open access
Chun-Hou Zheng, Lei Zhang, To-Yee Ng et al. · IEEE/ACM Transactions on Computational Biology and Bioinformatics · 2011
A reliable and accurate identification of the type of tumors is crucial to the proper treatment of cancers. In recent years, it has been shown that sparse representation (SR) by l1-norm minimization is robust to noise, outliers and even incomplete measurements, and SR has been successfully used for classification. This paper presents a new SR-based method for tumor classification using gene expression data. A set of metasamples are extracted from the training samples, and then an input testing sample is represented as the linear combination of these metasamples by l1-regularized least square method. Classification is achieved by using a discriminating function defined on the representation coefficients. Since l1-norm minimization leads to a sparse solution, the proposed method is called metasample-based SR classification (MSRC). Extensive experiments on publicly available gene expression data sets show that MSRC is efficient for tumor classification, achieving higher accuracy than many existing representative schemes.
Cited by 117
Yun Wang, Chen Xu, Peter J. Ramadge · 2013
The sparse representation of signals with respect to an over-complete dictionary has been of recent interest in a broad range of applications. One of the most used methods for obtaining sparse codes, the Lasso problem, becomes computationally costly for large dictionaries and this hinders the use of this approach to large-scale decision tasks. Recently, dictionary screening has been used to address this computational issue. In this spirit, we show how sequential Lasso screening can also facilitate faster completion of sparse representation decision tasks, such as classification, without affecting statistical accuracy. Moreover, the sequential screening process allows us to employ an early decision mechanism that can further accelerate classification, possibly at the cost of small decrease in accuracy.We demonstrate this empirically for several classification tasks. In particular, for clip-level music genre classification, using scattering features and a new voting scheme, we show that the proposed method yields improved clip classification accuracy and considerable computational speedup.
Cited by 4
Qi Zhu, Qingxiang Feng, Jiashuang Huang et al. · 2016
Sparse representation based classification has attracted much attention due to its robustness in the fields of biometrics, such as face recognition, palm-print recognition. SRC first constructs the linear representation model for the test sample, and then it classifies the test sample by comparing its coding error of each class. The coding error essentially can be viewed as the distance from sample to the subspace spanned by the samples from specific class. Therefore, the key to improve the classification performance of SRC is to enhance the subspace separability of these subspaces. In this paper, we introduce the difference subspace analysis into SRC, and propose difference subspace based SRC (DSSRC) for face recognition. Different from traditional dictionary learning based SRC methods, DSSRC focuses on maximizing the discriminability for the classes rather than the representation ability for the samples. Extensive experiments on the well-known image datasets demonstrate that the proposed DSSRC method is effective for face recognition.
Cited by 1