In this letter, a new sparse locality preserving projection (SLPP) algorithm is developed and applied to facial expression recognition. In comparison with the original locality preserving projection (LPP) algorithm, the presented SLPP algorithm is able to simultaneously find the intrinsic manifold of facial feature vectors and deal with facial feature selection. This is realized by the use of l1-norm regularization in the LPP objective function, which is directly formulated as a least squares regression pattern. We use two real facial expression databases (JAFFE and Ekman's POFA) to testify the proposed SLPP method and certain experiments show that the proposed SLPP approach respectively gains 77.60% and 82.29% on JAFFE and POFA database.

In this paper, we propose a representation method based on local spatial strokes for scene character recognition. High-level semantic information, namely co-occurrence of several strokes is incorporated by learning a sparse dictionary, which can further restrain noise brought by single stroke detectors. The encouraging results outperform state-of-the-art algorithms.

The label estimation technique provides a new way to design semi-supervised learning algorithms. If the labels of the unlabeled data can be estimated correctly, the semi-supervised methods can be replaced by the corresponding supervised versions. In this paper, we propose a novel semi-supervised learning algorithm, called Geodesic Weighted Sparse Representation (GWSR), to estimate the labels of the unlabeled data. First, the geodesic distance and geodesic weight are calculated. The geodesic weight is utilized to reconstruct the labeled samples. The Euclidean distance between the reconstructed labeled sample and the unlabeled sample equals the geodesic distance between the original labeled sample and the unlabeled sample. Then, the unlabeled samples are sparsely reconstructed and the sparse reconstruction weight is obtained by minimizing the L1-norm. Finally, the sparse reconstruction weight is utilized to estimate the labels of the unlabeled samples. Experiments on synthetic data and USPS hand-written digit database demonstrate the effectiveness of our method.

This paper presents a voice conversion (VC) technique for noisy environments based on a sparse representation of speech. Sparse representation-based VC using Non-negative matrix factorization (NMF) is employed for noise-added spectral conversion between different speakers. In our previous exemplar-based VC method, source exemplars and target exemplars are extracted from parallel training data, having the same texts uttered by the source and target speakers. The input source signal is represented using the source exemplars and their weights. Then, the converted speech is constructed from the target exemplars and the weights related to the source exemplars. However, this exemplar-based approach needs to hold all training exemplars (frames), and it requires high computation times to obtain the weights of the source exemplars. In this paper, we propose a framework to train the basis matrices of the source and target exemplars so that they have a common weight matrix. By using the basis matrices instead of the exemplars, the VC is performed with lower computation times than with the exemplar-based method. The effectiveness of this method was confirmed by comparing its effectiveness (in speaker conversion experiments using noise-added speech data) with that of an exemplar-based method and a conventional Gaussian mixture model (GMM)-based method.

We propose a cooperative compressed spectrum sensing scheme for correlated signals in wideband cognitive radio networks. In order to design a reconstruction algorithm which accurately recover the wideband signals from the compressed samples in low SNR (Signal-to-Noise Ratio) environments, we consider the multiple measurement vector model exploiting a sequence of input signals and propose a cooperative sparse Bayesian learning algorithm which models the temporal correlation of the input signals. Simulation results show that the proposed scheme outperforms existing compressed sensing algorithms for low SNRs.

In this letter, we present a radix-R regular interconnection pattern family of factorizations for the WHT-FFT with identical stage-to-stage interconnection pattern in a unified form, where R is any power of 2. This family of algorithms has identical sparse matrix factorization in each stage and can be implemented in a merged butterfly structure, which conduce to regular and efficient memory managing scalable to high radices. And in each stage, the butterflies with same twiddle factor set are aggregated together, which can reduce the twiddle factor evaluations or accesses to the lookup table. The kinds of factorization can also be extended to FFT, WHT and SCHT with identical stage-to-stage interconnection pattern.

This paper describes a new method for lossy audio signal compression via compressive sensing (CS). In this method, a structured shrinkage operator is employed to decompose the audio signal into three layers, with two sparse layers, tonal and transient, and additive noise, and then, both the tonal and transient layers are compressed using CS. Since the shrinkage operator is able to take into account the structure information of the coefficients in the transform domain, it is able to achieve a better sparse approximation of the audio signal than traditional methods do. In addition, we propose a sparsity allocation algorithm, which adjusts the sparsity between the two layers, thus improving the performance of CS. Experimental results demonstrated that the new method provided a better compression performance than conventional methods did.

A new type of the affine projection (AP) algorithms which incorporates the sparsity condition of a system is presented. To exploit the sparsity of the system, a weighted l1-norm regularization is imposed on the cost function of the AP algorithm. Minimizing the cost function with a subgradient calculus and choosing two distinct weightings for l1-norm, two stochastic gradient based sparsity regularized AP (SR-AP) algorithms are developed. Experimental results show that the SR-AP algorithms outperform the typical AP counterparts for identifying sparse systems.

In this Letter, a new face recognition approach based on curvelets and local ternary patterns (LTP) is proposed. First, we observe that the curvelet transform is a new anisotropic multi-resolution transform and can efficiently represent edge discontinuities in face images, and that the LTP operator is one of the best texture descriptors in terms of characterizing face image details. This motivated us to decompose the image using the curvelet transform, and extract the features in different frequency bands. As revealed by curvelet transform properties, the highest frequency band information represents the noisy information, so we directly drop it from feature selection. The lowest frequency band mainly contains coarse image information, and thus we deal with it more precisely to extract features as the face's details using LTP. The remaining frequency bands mainly represent edge information, and we normalize them for achieving explicit structure information. Then, all the extracted features are put together as the elementary feature set. With these features, we can reduce the features' dimension using PCA, and then use the sparse sensing technique for face recognition. Experiments on the Yale database, the extended Yale B database, and the CMU PIE database show the effectiveness of the proposed methods.

In this letter, we research the method of using face and gesture image sequences to deal with the video-based bimodal emotion recognition problem, in which both Harris plus cuboids spatio-temporal feature (HST) and sparse canonical correlation analysis (SCCA) fusion method are applied to this end. To efficaciously pick up the spatio-temporal features, we adopt the Harris 3D feature detector proposed by Laptev and Lindeberg to find the points from both face and gesture videos, and then apply the cuboids feature descriptor to extract the facial expression and gesture emotion features [1],[2]. To further extract the common emotion features from both facial expression feature set and gesture feature set, the SCCA method is applied and the extracted emotion features are used for the biomodal emotion classification, where the K-nearest neighbor classifier and the SVM classifier are respectively used for this purpose. We test this method on the biomodal face and body gesture (FABO) database and the experimental results demonstrate the better recognition accuracy compared with other methods.

Presented is a new measuring and reconstruction framework of Compressed Sensing (CS), aiming at reducing the measurements required to ensure faithful reconstruction. A sparse vector is segmented into sparser vectors. These new ones are then randomly sensed. For recovery, we reconstruct these vectors individually and assemble them to obtain the original signal. We show that the proposed scheme, referred to as SegOMP, yields higher probability of exact recovery in theory. It is finished with much smaller number of measurements to achieve a same reconstruction quality when compared to the canonical greedy algorithms. Extensive experiments verify the validity of the SegOMP and demonstrate its potentials.

This letter deals with a sparse signal recovery problem and proposes a new algorithm based on the iterative reweighted least squares (IRLS) algorithm. We assume that the non-zero values of a sparse signal is always greater than a given constant and modify the IRLS algorithm to satisfy this assumption. Numerical results show that the proposed algorithm recovers a sparse vector efficiently.

The sparse matrix operation, y ← y+AtAx, where A is a sparse matrix and x and y are dense vectors, is a widely used computing pattern in High Performance Computing (HPC) applications. The pattern poses challenge to efficient solutions because both a matrix and its transposed version are involved. An efficient sparse matrix format, Compressed Sparse Blocks (CSB), has been proposed to provide nearly the same performance for both Ax and Atx. We develop a multithreaded implementation for the CSB format and apply it to solve y ← y+AtAx. Experiments show that our technique outperforms the Compressed Sparse Row (CSR) based solution in POSKI by up to 2.5 fold on over 70% of benchmarking matrices.

Regularized forward selection is viewed as a method for obtaining a sparse representation in a nonparametric regression problem. In regularized forward selection, regression output is represented by a weighted sum of several significant basis functions that are selected from among a large number of candidates by using a greedy training procedure in terms of a regularized cost function and applying an appropriate model selection method. In this paper, we propose a model selection method in regularized forward selection. For the purpose, we focus on the reduction of a cost function, which is brought by appending a new basis function in a greedy training procedure. We first clarify a bias and variance decomposition of the cost reduction and then derive a probabilistic upper bound for the variance of the cost reduction under some conditions. The derived upper bound reflects an essential feature of the greedy training procedure; i.e., it selects a basis function which maximally reduces the cost function. We then propose a thresholding method for determining significant basis functions by applying the derived upper bound as a threshold level and effectively combining it with the leave-one-out cross validation method. Several numerical experiments show that generalization performance of the proposed method is comparable to that of the other methods while the number of basis functions selected by the proposed method is greatly smaller than by the other methods. We can therefore say that the proposed method is able to yield a sparse representation while keeping a relatively good generalization performance. Moreover, our method has an advantage that it is free from a selection of a regularization parameter.

Network tomography is a technique for estimating internal network characteristics from end-to-end measurements. In this paper, we focus on loss tomography, which is a network tomography problem for estimating link loss rates. We study a loss tomography problem to detect links with high link loss rates in network environments with dynamically changing link loss rates, and propose a window-based sequential loss tomography scheme. The loss tomography problem is formulated as an underdetermined linear inverse problem, where there are infinitely many candidates of the solution. In the proposed scheme, we use compressed sensing, which can solve the problem with a prior information that the solution is a sparse vector. Measurement nodes transmit probe packets on measurement paths established between them, and calculate packet loss rates of measurement paths (path loss rates) from probe packets received within a window. Measurement paths are classified into normal quality and low quality states according to the path loss rates. When a measurement node finds measurement paths in the low quality states, link loss rates are estimated by compressed sensing. Using simulation scenarios with a few link states changing dynamically from low to high link loss rates, we evaluate the performance of the proposed scheme.

This paper proposes an inter-prediction method for the upcoming video coding standard named HEVC (High Efficiency Video Coding). The HEVC offers an inter-prediction framework called local intensity compensation which represents a current block by a linear combination of some reference blocks. The proposed method calculates weight coefficients of the linear combination by using sparse representation. Experimental results show that the proposed method increases prediction accuracy in comparison with other methods.

In this paper, online sparse Volterra system identification is proposed. For that purpose, the conventional adaptive projection-based algorithm with weighted l1 balls (APWL1) is revisited for nonlinear system identification, whereby the linear-in-parameters nature of Volterra systems is utilized. Compared with sparsity-aware recursive least squares (RLS) based algorithms, requiring higher computational complexity and showing faster convergence and lower steady-state error due to their long memory in time-invariant cases, the proposed approach yields better tracking capability in time-varying cases due to short-term data dependence in updating the weight. Also, when N is the number of sparse Volterra kernels and q is the number of input vectors involved to update the weight, the proposed algorithm requires O(qN) multiplication complexity and O(Nlog 2N) sorting-operation complexity. Furthermore, sparsity-aware least mean-squares and affine projection based algorithms are also tested.

This paper deals with the problem of underdetermined blind source separation (BSS) where the number of sources is unknown. We propose a BSS approach that simultaneously estimates the number of sources, separates the sources based on the sparseness of speech, estimates the direction of arrival of each source, and performs permutation alignment. We confirmed experimentally that reasonably good separation was obtained with the present method without specifying the number of sources.

Near range microwave imaging systems have broad application prospects in the field of concealed weapon detection, biomedical imaging, nondestructive testing, etc. In this paper, the technique of optimized sparse antenna array is applied to near range microwave imaging, which can greatly reduce the complexity of imaging systems. In detail, the paper establishes three-dimensional sparse array imaging geometry and corresponding echo model, where the imaging geometry is formed by arranging optimized sparse antenna array in elevation, scanning in azimuth and transmitting broadband signals in range direction; and by analyzing the characteristics of near range imaging, that is, the maximum interval of transmitting and receiving elements is limited by the range from imaging system to targets, we propose the idea of piecewise sparse line array; secondly, by analyzing the convolution principle, we develop a method of arranging piecewise sparse array which can generate the same distribution of equivalent phase centers as filled antenna array; then, the paper deduces corresponding imaging algorithm; finally, the imaging geometry and corresponding algorithm proposed in this paper are investigated and verified via numerical simulations and near range imaging experiments.

For face recognition with a single training image per person, Collaborative Representation based Classification (CRC) has significantly less complexity than Extended Sparse Representation based Classification (ESRC). However, CRC gets lower recognition rates than ESRC. In order to combine the advantages of CRC and ESRC, we propose Extended Collaborative Representation based Classification (ECRC) for face recognition with a single training image per person. ECRC constructs an auxiliary intraclass variant dictionary to represent the possible variation between the testing and training images. Experimental results show that ECRC outperforms the compared methods in terms of both high recognition rates and low computation complexity.