China
Africa
Explore chapters and articles related to this topic
Vibration monitoring via spectro-temporal compressive sensing for wireless sensor networks
Published in Dan M. Frangopol, Hitoshi Furuta, Mitsuyoshi Akiyama, Dan M. Frangopol, Life-Cycle of Structural Systems, 2018
Although the CS methodology is relatively recent, the number of available solvers capable of dealing with the formulations of Equations (16),(18),(19) is already quite vast. Most of the methods derived for solving Equation (18) fall into the category of convex problem relaxation, i.e. replacement of the l1 norm with an equivalent formulation of the problem, or via greedy pursuit for obtaining a sparse and approximate solution. The most popular relaxation-based solvers include Gradient Projection for Sparse Reconstruction Figueiredo et al. (2007), Sparse reconstruction by separable approximation Wright et al. (2009), l1 regularised least squares problems (l1 ls) Kim et al. (2007), Spectral projected gradient (SPGL1) Birgin et al. (1999), Fast Iterative Soft-Thresholding Algorithm Beck and Teboulle (2009) and the Nesterov Algorithm (NESTA) Becker et al. (2011). Among the greedy-based algorithms the most popular ones involve CoSaMP Needell and Tropp (2009), Orthogonal Matching Pursuit Tropp (2004), and regularised Orthogonal Matching Pursuit Needell and Vershynin (2009).
Brain State Identification and Forecasting of Acute Pathology Using Unsupervised Fuzzy Clustering of EEG Temporal Patterns
Published in Horia-Nicolai Teodorescu, Abraham Kandel, Lakhmi C. Jain, FUZZY and NEURO-FUZZY SYSTEMS in MEDICINE, 2017
Mallat and Zhang [22] have introduced an algorithm, called matching pursuit, that decomposes any signal into a linear expansion of waveforms that are selected from a redundant dictionary of functions. These waveforms are chosen to match the signal structures. The matching pursuit algorithm can sometimes better isolate specific signal structures, which are not necessarily coherent with the wavelet form. At each iteration of the algorithm, a waveform that is best adapted to an approximate part of the signal is chosen. If a signal structure does not correlate well with any particular dictionary element, namely, a noise component, it is sub-decomposed into several elements and its information is diluted. Although matching pursuit is a nonlinear procedure, it does maintain an energy conservation, which guarantees its convergence [22].
Genetic Algorithm and BFOA-Based Iris and Palmprint Multimodal Biometric Digital Watermarking Models
Published in D. P. Acharjya, V. Santhi, Bio-Inspired Computing for Image and Video Processing, 2018
AlgorithmThe images are fed as input.The values of input image 1 and input image 2 are convolved.The resultant matrix is converted into sparse representation by squeezing out the zero elements.The sparse representation is further transformed into an image using orthogonal matching pursuit. The fused template is obtained.
Joint Dual-Structural Constrained and Non-negative Analysis Representation Learning for Pattern Classification
Published in Applied Artificial Intelligence, 2023
Kun Jiang, Lei Zhu, Qindong Sun
The solution of SDL model includes two basic tasks, i.e., sparse approximation and dictionary learning. On one hand, some algorithms, such as matching pursuit (M-P) (Davis, Mallat, and Avellaneda 1997; Mallat and Zhang 1993), basis pursuit (BP) (Chen, Donoho, and Saunders 2001) and shrinkage method (Hyvärinen 1999), have been well developed to find a sparse solution. On the other hand, dictionary learning is dedicated to search an optimal signal space to support the attribution of sparse vector under a certain measure. There exist a variety of numerical algorithms presented to achieve this objective, e.g., method of optimal directions (MOD) (Engan, Aase, and Hakon Husoy 1999) and K-singular value decomposition (K-SVD) (Aharon, Elad, and Bruckstein 2006). K-SVD method learns an overcomplete dictionary from training samples by updating K dictionary atoms and representation coefficients iteratively with the SVD algorithm under a predefined sparse threshold for non-zero elements in each coefficient (Aharon, Elad, and Bruckstein 2006).
Scalable level-wise screening experiments using locating arrays
Published in Journal of Quality Technology, 2023
Yasmeen Akhtar, Fan Zhang, Charles J. Colbourn, John Stufken, Violet R. Syrotiuk
The proposed level-wise screening method has two steps. First, a breadth-first search (BFS) algorithm is developed to identify a user-specified number of level-wise models that are the “best” explanations of a response using orthogonal matching pursuit (OMP; Davis, Mallat, and Avellaneda 1997), which is widely used in signal processing (Tropp and Gilbert 2007) to recover sparse signals. A matching pursuit is a greedy algorithm that progressively refines an approximation of an optimization problem with an iterative procedure instead of solving it optimally. The vector selected at each iteration by the matching pursuit algorithm is generally not orthogonal to the previously selected vectors. In OMP, the approximations are refined by orthogonalizing the directions of projection.
Power Quality Disturbances Classification Using Compressive Sensing and Maximum Likelihood
Published in IETE Technical Review, 2018
Hui Liu, Fida Hussain, Yue Shen
CS [31–33] is a novel signal and image processing theory for acquiring and reconstructing a sparse signal from a small set of non-adaptive, linear measurements. Assume be a signal, given an orthonormal basis matrix whose basis are columns, can be represented as , where is an column sparse vector with only non-zero elements, that is . The measurement process can be calculated as where is a column vector of the compressive measurement, is an () measurement matrix and incoherent with , and sensing matrix should satisfy the Restricted Isometry Property. Many algorithms have been used for the recovery of signals such as basis pursuit, matching pursuit, and orthogonal matching pursuit.