A relatively fast pursuit algorithm in face recognition is proposed, compared to existing pursuit algorithms. More stopping rules have been put forward to solve the problem of slow response of OMP, which can fully dev...
详细信息
A relatively fast pursuit algorithm in face recognition is proposed, compared to existing pursuit algorithms. More stopping rules have been put forward to solve the problem of slow response of OMP, which can fully develop the superiority of pursuit algorithm - avoiding to process useless information in the training dictionary. For the test samples that are affected by partial occlusion, corruption, and facial disguise, recognition rates of most algorithms fall rapidly. The robust version of this algorithm can identify these samples automatically and process them accordingly. The recognition rates on ORL database, Yale database, and FERET database are 95.5%, 93.87%, and 92.29%, respectively. The recognition performance under various levels of occlusion and corruption is also experimentally proved to be significantly enhanced.
Many numerical methods for computing radiation exchange in enclosures are based on the computation of direct exchange areas (DEAs) and total exchange areas (TEAs). Excessively long computation times can be associated ...
详细信息
Many numerical methods for computing radiation exchange in enclosures are based on the computation of direct exchange areas (DEAs) and total exchange areas (TEAs). Excessively long computation times can be associated with TEAs computation. Among the best performing methods, the plating algorithm (PA) computes TEAs from DEAs based on a set of simple recursive equations. An efficient CPU and GPU parallelization of the PA are represented. Nevertheless, PA computation complexity is O(N-3). A novel formulation, the nonrecursive plating algorithm (NRPA), is introduced. It allows the computation of TEAs with a single nonrecursive step. Its equations are formulated by identification to the PA equations giving TEAs from DEAs, requiring one simple assumption. The NRPA is then written in matrix form as mainly a square matrix multiplication operation. Based on advancement in matrix multiplication computation, the NRPA complexity is proven to be O(N-2.38) for the number of multiplications. CPU and GPU NRPA are implemented based on the optimized linear algebra library BLAS for CPU and cuBLAS for GPU CUDA programs. NRPA is found to highly outperform PA in both CPU and GPU computation times. Finally, a test enclosure is considered and serves to validate the accuracy of the NRPA by comparison to the PA.
Prediction of liquidliquid phase equilibria in reacting systems is important in many applications such as reactive extraction. This problem poses several numerical challenges. These systems are governed by highly nonl...
详细信息
Prediction of liquidliquid phase equilibria in reacting systems is important in many applications such as reactive extraction. This problem poses several numerical challenges. These systems are governed by highly nonlinear algebraic equations and are plagued by the issue of nonconvergence of iterative algorithms. They exhibit strong sensitivity to the choice of the initial values or starting guesses for the variables. The nonconvergence stems primarily from the wide range of concentrations of various species at equilibrium. In this work, a new methodology for predicting thermodynamic equilibria of multiphase reacting systems is proposed. The mathematical formulation and solution are based on the use of the logarithms of the concentrations as the dependent variables. The proposed algorithm shows rapid convergence even when the initial guesses are far from equilibrium. The efficiency of the method is demonstrated by predicting the equilibrium concentrations of species in three systems of varying degrees of complexity.
This study deals with a NN (neural-network)-based control algorithm of a grid interfaced SPV (solar photovoltaic) generating system. The proposed grid interfaced SPV generating system utilises a NN control algorithm-b...
详细信息
This study deals with a NN (neural-network)-based control algorithm of a grid interfaced SPV (solar photovoltaic) generating system. The proposed grid interfaced SPV generating system utilises a NN control algorithm-based on the LMS (least mean-square), known as Adaline (adaptive linear element) to estimate reference grid currents. A DC-DC boost converter is used for achieving the maximum power point tracking between SPV and DC bus of four-leg VSC (voltage source converter) interfaced to a three-phase, four-wire distribution system. The four-leg VSC of SPV generating system is also used for the compensation of the reactive power for zero voltage regulation or for power factor correction along with load balancing, elimination of load harmonics currents and mitigation of neutral current at PCC (point of common coupling) in three-phase four-wire distribution system. The DC bus of VSC is supported by a capacitor which is fed by SPV energy through a DC-DC boost converter. A laboratory prototype of proposed grid interfaced SPV generating system is developed to validate its developed model and the NN-based control algorithm.
We examine several algorithms for tracking a handheld wand in a 3D virtual reality system: extended Kalman filters (EKFs), interacting multiple models (IMMs), and support vector machines (SVMs). The IMMs consist of se...
详细信息
We examine several algorithms for tracking a handheld wand in a 3D virtual reality system: extended Kalman filters (EKFs), interacting multiple models (IMMs), and support vector machines (SVMs). The IMMs consist of several EKF models, each of which is tuned for one particular type of user motion. For determining the types of motion, we compare hand-created rules with an automatic clustering algorithm, with mixed results. The mode-specific EKFs within the IMM are more accurate than one overall EKF. However, the IMM is comparable to a single EKF, because of the overhead of predicting the current component EKF. SVMs with a one-frame lookahead perform the best, cutting the error in half. Aside from those SVMs, different model types were best for the different dimensions of tracking (x, y, z, and rotation).
We investigated the possible frequency dependence of the moment tensor of large earthquakes by performing W phase inversions using teleseismic data and equally-spaced narrow, overlapping frequency bands. We investigat...
详细信息
We investigated the possible frequency dependence of the moment tensor of large earthquakes by performing W phase inversions using teleseismic data and equally-spaced narrow, overlapping frequency bands. We investigated frequencies from 0.6 to 3.8 mHz. Our focus was on the variation with frequency of the scalar moment, the amount of non-double-couple, and the focal mechanism. We applied this technique to 30 major events in the period 1994-2013 and used the results to detect source complexity. Based on the results, we classed them into three groups according to the variability of the source parameters with frequency: simple, complex and intermediate. Twelve of these events fell into the simple category: Bolivia-1994, Kuril-1994, Sanriku-1994, Antofagasta-1995, Andreanoff-1996, Peru-2001, Sumatra-2004, Sumatra-2005, Tonga-2006, Sumatra-2007, Japan-2011, and the recent Sea of Okhotsk-2013. Seven exhibited significant complexity: Balleny-1998, Sumatra-2000, Indian Ocean-2000, Macquarie Island-2004, Sichuan-2008, and Samoa-2009. The remaining 11 events showed a moderate degree of complexity. Here, we discuss the results of this study in light of independent observations of source complexity, made by various investigators.
In this paper, we consider the speech enhancement problem in a moving car through a blind source separation (BSS) scheme involving two spaced microphones. The forward and backward blind source separation structures of...
详细信息
In this paper, we consider the speech enhancement problem in a moving car through a blind source separation (BSS) scheme involving two spaced microphones. The forward and backward blind source separation structures often use manual voice activity detector (MVAD) systems to control the adaptation of the separating adaptive filters. In this paper, we propose two new automatic voice activity detector (AVAD) systems that allow adapting the original forward and backward BSS structures automatically. The proposed AVAD systems are based on the use of the forward BSS structure to estimate the optimal values of the separating adaptive filters step-sizes. Moreover, the new proposed algorithms are stable and could be used even in very noisy conditions. Intensive experiments are carried out with these two new proposed algorithms to validate their good performances in speech enhancement and noise reduction applications. The presented experiments are based on the system mismatch, the cepstral distance and the output signal-to-noise ratio criteria evaluations. The obtained results show the good performances of the proposed algorithms in comparison with their original versions, where manual VAD systems are used. (C) 2014 Elsevier Ltd. All rights reserved.
This paper develops new methods for vision-based satellite attitude control aimed at space-based optical tracking of resident space objects (RSOs). An Earth-orbiting chaser satellite equipped with either one or two bo...
详细信息
This paper develops new methods for vision-based satellite attitude control aimed at space-based optical tracking of resident space objects (RSOs). An Earth-orbiting chaser satellite equipped with either one or two body-fixed cameras can successfully track an RSO provided that the target is kept within the camera field of view. Because the cameras are body fixed, the attitude of the satellite needs to be controlled to maintain target lock. Novel vision-based control algorithms are developed to align the chaser camera's optical axis with the chaser-target line of sight. Two control architectures are presented for the cases of monocular and stereoscopic vision. In the case of the monocular architecture, relative chaser-target acceleration information is not available. Moreover, in both cases unknown perturbations can impair the tracking performance. To increase the tracking algorithm's robustness to these effects, a variable structure attitude control technique is employed. The stability of the developed control laws are substantiated based on Lyapunov's direct method and demonstrated using Monte Carlo simulations. The results clearly show that using stereoscopic vision yields faster target tracking, increased robustness to noise and field-of-view limits, and reduced fuel consumption compared with monocular vision-based attitude tracking.
The problem of reconstruction of inclusions in rough diamonds occupies an important place in the entire technological process of extraction and cutting precious stones. Inclusions are foreign objects that are naturall...
详细信息
The problem of reconstruction of inclusions in rough diamonds occupies an important place in the entire technological process of extraction and cutting precious stones. Inclusions are foreign objects that are naturally formed in raw stock during the formation of diamond (see Fig. 1). The algorithm proposed in the present paper performs a full reconstruction of three-dimensional models of inclusions from photo images. Three-dimensional models of inclusions are based on a voxel representation. The algorithm performs the coloring of voxels and the segmentation of a voxel grid and constructs polygonal models of inclusions. Experiments carried out on real and synthetic data show that the algorithm can reconstruct three-dimensional models of medium- and large-size inclusions, which may serve as a first approximation for the methods of refinement of the shape of inclusions.
The problem of stabilization of rigid bodies has received a great deal of attention for many years. People have developed a variety of feedback control laws to meet their design requirements and have formulated variou...
详细信息
The problem of stabilization of rigid bodies has received a great deal of attention for many years. People have developed a variety of feedback control laws to meet their design requirements and have formulated various but mostly open loop numerical algorithms for the dynamics of the corresponding closed loop systems. Since the conserved quantities such as energy, momentum, and symmetry play an important role in the dynamics, we investigate the conserved quantities for the closed loop control systems which formally or asymptotically stabilize rigid body rotation and modify the open loop numerical algorithms so that they preserve these important properties. Using several examples, the authors first use the open loop algorithm to simulate the tumbling rigid body actions and then use the resulting closed loop one to stabilize them.
暂无评论