Volumen 24 (2014): Heft 2 (June 2014)
Signals and Systems (special section, pp. 233-312), Ryszard Makowski and Jan Zarzycki (Eds.)

Nonnegative Matrix Factorization (NMF) is an important tool in data spectral analysis. However, when a mixing matrix or sources are not sufficiently sparse, NMF of an observation matrix is not unique. Many numerical optimization algorithms, which assure fast convergence for specific problems, may easily get stuck into unfavorable local minima of an objective function, resulting in very low performance. In this paper, we discuss the Tikhonov regularized version of the Fast Combinatorial NonNegative Least Squares (FC-NNLS) algorithm (proposed by Benthem and Keenan in 2004), where the regularization parameter starts from a large value and decreases gradually with iterations. A geometrical analysis and justification of this approach are presented. The numerical experiments, carried out for various benchmarks of spectral signals, demonstrate that this kind of regularization, when applied to the FC-NNLS algorithm, is essential to obtain good performance.

In this work, the problem of coil design is studied. It is assumed that the structure of the coil is known (i.e., the positions of simple circular coils are fixed) and the problem is to find current distribution to obtain the required magnetic field in a given region. The unconstrained version of the problem (arbitrary currents are allowed) can be formulated as a Least-SQuares (LSQ) problem. However, the results obtained by solving the LSQ problem are usually useless from the application point of view. Moreover, for higher dimensions the problem is ill-conditioned. To overcome these difficulties, a regularization term is sometimes added to the cost function, in order to make the solution smoother. The regularization technique, however, produces suboptimal solutions. In this work, we propose to solve the problem under study using the constrained Quadratic Programming (QP) method. The methods are compared in terms of the quality of the magnetic field obtained, and the power of the designed coil. Several 1D and 2D examples are considered. It is shown that for the same value of the maximum current the QP method provides solutions with a higher quality magnetic field than the regularization method.

Speech segmentation is an essential stage in designing automatic speech recognition systems and one can find several algorithms proposed in the literature. It is a difficult problem, as speech is immensely variable. The aim of the authors’ studies was to design an algorithm that could be employed at the stage of automatic speech recognition. This would make it possible to avoid some problems related to speech signal parametrization. Posing the problem in such a way requires the algorithm to be capable of working in real time. The only such algorithm was proposed by Tyagi et al., (2006), and it is a modified version of Brandt’s algorithm. The article presents a new algorithm for unsupervised automatic speech signal segmentation. It performs segmentation without access to information about the phonetic content of the utterances, relying exclusively on second-order statistics of a speech signal. The starting point for the proposed method is time-varying Schur coefficients of an innovation adaptive filter. The Schur algorithm is known to be fast, precise, stable and capable of rapidly tracking changes in second order signal statistics. A transfer from one phoneme to another in the speech signal always indicates a change in signal statistics caused by vocal track changes. In order to allow for the properties of human hearing, detection of inter-phoneme boundaries is performed based on statistics defined on the mel spectrum determined from the reflection coefficients. The paper presents the structure of the algorithm, defines its properties, lists parameter values, describes detection efficiency results, and compares them with those for another algorithm. The obtained segmentation results, are satisfactory.

The paper presents the state of the art and advancement of artificial intelligence methods in analog systems diagnostics. Firstly, the diagnostic domain is introduced and its problems explained. Then, computational intelligence approaches usable for fault detection and identification are reviewed. Particular groups of methods are presented in detail, explaining their usefulness and drawbacks. Examples, such as the induction motor or the electronic filter, are provided to show the applicability of the presented approaches for monitoring the state of analog objects from engineering domains. The discussion section reviews the presented approaches, their future prospects and problems to be solved.

The paper concerns the problem of Boolean satisfiability checking, which is recognized as one of the most important issues in the field of modern digital electronic system verification and design. The paper analyzes different strategies and scenarios of the proving process, and presents a modified and extended version of the author’s FUDASAT algorithm. The original FUDASAT methodology is an intuitive approach that employs a commonsense reasoning methodology. The main objective of the work is to investigate the SAT-solving process and try to formulate a set of rules controlling the reasoning process of the FUDASAT inference engine. In comparison with the author’s previous works, the paper introduces new mechanisms: hypergraph analysis, multiple variable assignments and search space pruning algorithms. The approach considers only 3-SAT class functions, although a generalization of the method is discussed as well. The presented approach has been tested on various benchmarks and compared with the original pure FUDASAT algorithm as well as with other algorithms known from the literature. Finally, the benefits of the proposed SAT solving technique are summarized.

Hydrocephalus is a pathological condition of the central nervous system which often affects neonates and young children. It manifests itself as an abnormal accumulation of cerebrospinal fluid within the ventricular system of the brain with its subsequent progression. One of the most important diagnostic methods of identifying hydrocephalus is Computer Tomography (CT). The enlarged ventricular system is clearly visible on CT scans. However, the assessment of the disease progress usually relies on the radiologist’s judgment and manual measurements, which are subjective, cumbersome and have limited accuracy. Therefore, this paper regards the problem of semi-automatic assessment of hydrocephalus using image processing and analysis algorithms. In particular, automated determination of popular indices of the disease progress is considered. Algorithms for the detection, semi-automatic segmentation and numerical description of the lesion are proposed. Specifically, the disease progress is determined using shape analysis algorithms. Numerical results provided by the introduced methods are presented and compared with those calculated manually by a radiologist and a trained operator. The comparison proves the correctness of the introduced approach.

Disturbance modeling and design of state estimators for offset-free Model Predictive Control (MPC) with linear state-space process models is considered in the paper for deterministic constant-type external and internal disturbances (modeling errors). The application and importance of constant state disturbance prediction in the state-space MPC controller design is presented. In the case with a measured state, this leads to the control structure without disturbance state observers. In the case with an unmeasured state, a new, simpler MPC controller-observer structure is proposed, with observation of a pure process state only. The structure is not only simpler, but also with less restrictive applicability conditions than the conventional approach with extended process-and-disturbances state estimation. Theoretical analysis of the proposed structure is provided. The design approach is also applied to the case with an augmented state-space model in complete velocity form. The results are illustrated on a 2×2 example process problem.

This paper is concerned with the problem of designing a robust modified repetitive-control system with a dynamic output feedback controller for a class of strictly proper plants. Employing the continuous lifting technique, a continuous-discrete two-dimensional (2D) model is built that accurately describes the features of repetitive control. The 2D control input contains the direct sum of the effects of control and learning, which allows us to adjust control and learning preferentially. The singular-value decomposition of the output matrix and Lyapunov stability theory are used to derive an asymptotic stability condition based on a Linear Matrix Inequality (LMI). Two tuning parameters in the LMI manipulate the preferential adjustment of control and learning. A numerical example illustrates the tuning procedure and demonstrates the effectiveness of the method.

A minimum energy control problem for fractional positive continuous-time linear systems with bounded inputs is formulated and solved. Sufficient conditions for the existence of a solution to the problem are established. A procedure for solving the problem is proposed and illustrated with a numerical example.

This paper deals with stability analysis of hybrid systems. Various stability concepts related to hybrid systems are introduced. The paper advocates a local analysis. It involves the equivalence relation generated by reset maps of a hybrid system. To establish a tangible method for stability analysis, we introduce the notion of a chart, which locally reduces the complexity of the hybrid system. In a chart, a hybrid system is particularly simple and can be analyzed with the use of methods borrowed from the theory of differential inclusions. Thus, the main contribution of this paper is to show how stability of a hybrid system can be reduced to a specialization of the well established stability theory of differential inclusions. A number of examples illustrate the concepts introduced in the paper.

The paper presents the training problem of a set of neural nets to obtain a (gain-scheduling, adaptive) multivariable neural controller for control of a nonlinear MIMO dynamic process represented by a mathematical model of Low-Frequency (LF) motions of a drillship over the drilling point at the sea bottom. The designed neural controller contains a set of neural nets that determine values of its parameters chosen on the basis of two measured auxiliary signals. These are the ship’s current forward speed measured with respect to water and the systematically calculated difference between the course angle and the sea current (yaw angle). Four different methods for synthesis of multivariable modal controllers are used to obtain source data for training the neural controller with parameters reproduced by neural networks. Neural networks are designed on the basis of 3650 modal controllers obtained with the use of the pole placement technique after having linearized the model of LF motions made by the vessel at its nominal operating points in steady states that are dependent on the specified yaw angle and the sea current velocity. The final part of the paper includes simulation results of system operation with a neural controller along with conclusions and final remarks.

In this paper, we address the pursuit-evasion problem of tracking an Omnidirectional Agent (OA) at a bounded variable distance using a Differential Drive Robot (DDR), in an Euclidean plane without obstacles. We assume that both players have bounded speeds, and that the DDR is faster than the evader, but due to its nonholonomic constraints it cannot change its motion direction instantaneously. Only a purely kinematic problem is considered, and any effect due to dynamic constraints (e.g., acceleration bounds) is neglected. We provide a criterion for partitioning the configuration space of the problem into two regions, so that in one of them the DDR is able to control the system, in the sense that, by applying a specific strategy (also provided), the DDR can achieve any inter-agent distance (within an error bound), regardless of the actions taken by the OA. Particular applications of these results include the capture of the OA by the DDR and maintaining surveillance of the OA at a bounded variable distance.

In this paper, a singularly perturbed system of reaction–diffusion Boundary Value Problems (BVPs) is examined. To solve such a type of problems, a Modified Initial Value Technique (MIVT) is proposed on an appropriate piecewise uniform Shishkin mesh. The MIVT is shown to be of second order convergent (up to a logarithmic factor). Numerical results are presented which are in agreement with the theoretical results.

Automated Incident Detection (AID) is an important part of Advanced Traffic Management and Information Systems (ATMISs). An automated incident detection system can effectively provide information on an incident, which can help initiate the required measure to reduce the influence of the incident. To accurately detect incidents in expressways, a Support Vector Machine (SVM) is used in this paper. Since the selection of optimal parameters for the SVM can improve prediction accuracy, the tabu search algorithm is employed to optimize the SVM parameters. The proposed model is evaluated with data for two freeways in China. The results show that the tabu search algorithm can effectively provide better parameter values for the SVM, and SVM models outperform Artificial Neural Networks (ANNs) in freeway incident detection.

With the aim to better preserve sharp edges and important structure features in the recovered image, this article researches an improved adaptive total variation regularization and H⁻¹ norm fidelity based strategy for image decomposition and restoration. Computationally, for minimizing the proposed energy functional, we investigate an efficient numerical algorithm—the split Bregman method, and briefly prove its convergence. In addition, comparisons are also made with the classical OSV (Osher–Sole–Vese) model (Osher et al., 2003) and the TV-Gabor model (Aujol et al., 2006), in terms of the edge-preserving capability and the recovered results. Numerical experiments markedly demonstrate that our novel scheme yields significantly better outcomes in image decomposition and denoising than the existing models.

This paper presents a new set of bivariate discrete orthogonal moments which are based on bivariate Hahn polynomials with non-separable basis. The polynomials are scaled to ensure numerical stability. Their computational aspects are discussed in detail. The principle of parameter selection is established by analyzing several plots of polynomials with different kinds of parameters. Appropriate parameters of binary images and a grayscale image are obtained through experimental results. The performance of the proposed moments in describing images is investigated through several image reconstruction experiments, including noisy and noise-free conditions. Comparisons with existing discrete orthogonal moments are also presented. The experimental results show that the proposed moments outperform slightly separable Hahn moments for higher orders.

A packet buffer limited to a fixed number of packets (regardless of their lengths) is considered. The buffer is described as a finite FIFO queuing system fed by a Markovian Arrival Process (MAP) with service times forming a Semi-Markov (SM) process (MAP /SM /1/b in Kendall’s notation). Such assumptions allow us to obtain new analytical results for the queuing characteristics of the buffer. In the paper, the following are considered: the time to fill the buffer, the local loss intensity, the loss ratio, and the total number of losses in a given time interval. Predictions of the proposed model are much closer to the trace-driven simulation results compared with the prediction of the MAP /G/1/b model.