Noise reduction is an open problem and has received considerable attention in the literature for several decades. Over the last two decades, wavelet based methods have been applied to the problem of noise reduction and have been shown to outperform the traditional Wiener filter, Median filter, and modified Lee filter in terms of root mean squared error (MSE), peak signal noise ratio (PSNR) and other evaluation methods. In this research, two approaches for the development of high performance algorithms for de-noising are proposed, both based on soft computing tools, such as fuzzy logic, neural networks, and genetic algorithms. First, an improved additive noise reduction method for digital grey scale nature images, which uses an interval type-2 fuzzy logic system to shrink wavelet coefficients, is proposed. This method is an extension of a recently published approach for additive noise reduction using a type-1 fuzzy logic system based wavelet shrinkage. Unlike the type-1 fuzzy logic system based wavelet shrinkage method, the proposed approach employs a thresholding filter to adjust the wavelet coefficients according to the linguistic uncertainty in neighborhood values, inter-scale dependencies and intra-scale correlations of wavelet coefficients at different resolutions by exploiting the interval type-2 fuzzy set theory. Experimental results show that the proposed approach can efficiently and rapidly remove additive noise from digital grey scale images. Objective analysis and visual observations show that the proposed approach outperforms current fuzzy non-wavelet methods and fuzzy wavelet based methods, and is comparable with some recent but more complex wavelet methods, such as Hidden Markov Model based additive noise de-noising method. The main differences between the proposed approach and other wavelet shrinkage based approaches and the main improvements of the proposed approach are also illustrated in this thesis. Second, another improved method of additive noise reduction is also proposed. The method is based on fusing the results of different filters using a Fuzzy Neural Network (FNN). The proposed method combines the advantages of these filters and has outstanding ability of smoothing out additive noise while preserving details of an image (e.g. edges and lines) effectively. A Genetic Algorithm (GA) is applied to choose the optimal parameters of the FNN. The experimental results show that the proposed method is powerful for removing noise from natural images, and the MSE of this approach is less, and the PSNR of is higher, than that of any individual filters which are used for fusion. Finally, the two proposed approaches are compared with each other from different point of views, such as objective analysis in terms of mean squared error(MSE), peak signal to noise ratio (PSNR), image quality index (IQI) based on quality assessment of distorted images, and Information Theoretic Criterion (ITC) based on a human vision model, computational cost, universality, and human observation. The results show that the proposed FNN based algorithm optimized by GA has the best performance among all testing approaches. Important considerations for these proposed approaches and future work are discussed.

This thesis develops strain and kinetic energy functions and a finite beam element useful for analyzing curved beams which go through large deflections, such as a hockey stick being swung and bent substantially as it hits the ice. The resulting beam model is demonstrated to be rotation invariant and capable of computing the correct strain energy and reaction forces for a specified deformation. A method is also described by which the model could be used to perform static or dynamic simulations of a beam.

Exploring a set of unlabeled data to extract the similar clusters, known as data clustering, is an appealing problem in machine learning. In other words, data clustering organizes the underlying data into different groups using a notion of similarity between patterns. A new approach to solve the data clustering problem based on multiple cooperative swarms is introduced. The proposed approach is inspired by the social swarming behavior of biological bird flocks which search for food situated in several places. The proposed approach is composed of two main phases, namely, initialization and exploitation. In the initialization phase, the aim is to distribute the search space among several swarms. That is, a part of the search space is assigned to each swarm in this phase. In the exploitation phase, each swarm searches for the center of its associated cluster while cooperating with other swarms. The search proceeds to converge to a near-optimal solution. As compared to the single swarm clustering approach, the proposed multiple cooperative swarms provide better solutions in terms of fitness function measure for the cluster centers, as the dimensionality of data and number of clusters increase. The multiple cooperative swarms clustering approach assumes that the number of clusters is known a priori. The notion of stability analysis is proposed to extract the number of clusters for the underlying data using multiple cooperative swarms. The mathematical explanations demonstrating why the proposed approach leads to more stable and robust results than those of the single swarm clustering are also provided. Application of the proposed multiple cooperative swarms clustering is considered for one of the most challenging problems in speech recognition: phoneme recognition. The proposed approach is used to decompose the recognition task into a number of subtasks or modules. Each module involves a set of similar phonemes known as a phoneme family. Basically, the goal is to obtain the best solution for phoneme families using the proposed multiple cooperative swarms clustering. The experiments using the standard TIMIT corpus indicate that using the proposed clustering approach boosts the accuracy of the modular approach for phoneme recognition considerably.

Machine intelligence techniques contribute to solving real-world problems. Reinforcement learning (RL) is one of the machine intelligence techniques with several characteristics that make it suitable for the applications, for which the model of the environment is not available to the agent. In real-world applications, intelligent agents generally face a very large state space which limits the usability of reinforcement learning. The condition for convergence of reinforcement learning implies that each state-action pair must be visited infinite times, a condition which can be considered impossible to be satisfied in many practical situations. The goal of this work is to propose a class of new techniques to overcome this problem for off-policy, step-by-step (incremental) and model-free reinforcement learning with discrete state and action space. The focus of this research is using the design characteristics of RL agent to improve its performance regarding the running time while maintaining an acceptable level of accuracy. One way of improving the performance of the intelligent agents is using the model of environment. In this work, a special type of knowledge about the agent actions is employed to improve its performance because in many applications the model of environment may only be known partially or not at all. The concept of opposition is employed in the framework of reinforcement learning to achieve this goal. One of the components of RL agent is the action. For each action we define its associate opposite action. The actions and opposite actions are implemented in the framework of reinforcement learning to update the value function resulting in a faster convergence. At the beginning of this research the concept of opposition is incorporated in the components of reinforcement learning, states, actions, and reinforcement signal which results in introduction of the oppositional target domain estimation algorithm, OTE. OTE reduces the search and navigation area and accelerates the speed of search for a target. The OTE algorithm is limited to the applications, in which the model of the environment is provided for the agent. Hence, further investigation is conducted to extend the concept of opposition to the model-free reinforcement learning algorithms. This extension contributes to the generating of several algorithms based on using the concept of opposition for Q(lambda) technique. The design of reinforcement learning agent depends on the application. The emphasize of this research is on the characteristics of the actions. Hence, the primary challenge of this work is design and incorporation of the opposite actions in the framework of RL agents. In this research, three different applications, namely grid navigation, elevator control problem, and image thresholding are implemented to address this challenge in context of different applications. The design challenges and some solutions to overcome the problems and improve the algorithms are also investigated. The opposition-based Q(lambda) algorithms are tested for the applications mentioned earlier. The general idea behind the opposition-based Q(lambda) algorithms is that in Q-value updating, the agent updates the value of an action in a given state. Hence, if the agent knows the value of opposite action then instead of one value, the agent can update two Q-values at the same time without taking its corresponding opposite action causing an explicit transition to opposite state. If the agent knows both values of action and its opposite action for a given state, then it can update two Q-values. This accelerates the learning process in general and the exploration phase in particular. Several algorithms are outlined in this work. The OQ(lambda) will be introduced to accelerate Q(lambda) algorithm in discrete state spaces. The NOQ(lambda) method is an extension of OQ(lambda) to operate in a broader range of non-deterministic environments. The update of the opposition trace in OQ(lambda) depends on the next state of the opposite action (which generally is not taken by the agent). This limits the usability of this technique to the deterministic environments because the next state should be known to the agent. NOQ(lambda) will be presented to update the opposition trace independent of knowing the next state for the opposite action. The results show the improvement of the performance in terms of running time for the proposed algorithms comparing to the standard Q(lambda) technique.

In this thesis, modeling and optimization in the field of storage management under stochastic condition will be investigated using two different methodologies: Simulation Optimization Techniques (SOT), which are usually categorized in the area of Reinforcement Learning (RL), and Nonlinear Modeling Techniques (NMT). For the first set of methods, simulation plays a fundamental role in evaluating the control policy: learning techniques are used to deliver sub-optimal policies at the end of a learning process. These iterative methods use the interaction of agents with the stochastic environment through taking actions and observing different states. To converge to the steady-state condition where policies and value functions do not change significantly with the continuation of the learning process, all or most important states must be visited sufficiently. This might be prohibitively time-consuming for large-scale problems. To make these techniques more efficient both in terms of computation time and robust optimal policies, the idea of Opposition-Based Learning (OBL-Type I and Type II) is employed to modify/extend popular RL techniques including Q-Learning, Q(λ), sarsa, and sarsa(λ). Several new algorithms are developed using this idea. It is also illustrated that, function approximation techniques such as neural networks can contribute to the process of learning. The state-of-the-art implementations usually consider the maximization of expected value of accumulated reward. Extending these techniques to consider risk and solving some well-known control problems are important contributions of this thesis. Furthermore, the new nonlinear modeling for reservoir management using indicator functions and randomized policy introduced by Fletcher and Ponnambalam, is extended to stochastic releases in multi-reservoir systems. In this extension, two different approaches for defining the release policies are proposed. In addition, the main restriction of considering the normal distribution for inflow is relaxed by using a beta-equivalent general distribution. A five-reservoir case study from India is used to demonstrate the benefits of these new developments. Using a warehouse management problem as an example, application of the proposed method to other storage management problems is outlined.

Objectives: Based on the analysis of electromyographic (EMG) data muscles are often characterized as normal or affected by a neuromuscular disease process. A clinical decision support system (CDSS) for the electrophysiological characterization of muscles by analyzing motor unit potentials (MUPs) was developed to assist physicians and researchers with the diagnosis, treatment & management of neuromuscular disorders and analyzed against criteria for use in a clinical setting. Methods: Quantitative MUP data extracted from various muscles from control subjects and patients from a number of clinics was used to compare the sensitivity, specificity, and accuracy of a number of different clinical decision support methods. The CDSS developed in this work known as AMC-PD has three components: MUP characterization using Pattern Discovery (PD), muscle characterization by taking the average of MUP characterizations and calibrated muscle characterizations. Results: The results demonstrated that AMC-PD achieved higher accuracy than conventional means and outlier analysis. Duration, thickness and number of turns were the most discriminative MUP features for characterizing the muscles studied in this work. Conclusions: AMC-PD achieved higher accuracy than conventional means and outlier analysis. Muscle characterization performed using AMC-PD can facilitate the determination of “possible”, “probable”, or “definite” levels of disease whereas the conventional means and outlier methods can only provide a dichotomous “normal” or “abnormal” decision. Therefore, AMC-PD can be directly used to support clinical decisions related to initial diagnosis as well as treatment and management over time. Decisions are based on facts and not impressions giving electromyography a more reliable role in the diagnosis, management, and treatment of neuromuscular disorders. AMC-PD based calibrated muscle characterization can help make electrophysiological examinations more accurate and objective.

Epilepsy is a serious neurological disorder characterized by recurrent unprovoked seizures due to abnormal or excessive neuronal activity in the brain. An estimated 50 million people around the world suffer from this condition, and it is classified as the second most serious neurological disease known to humanity, after stroke. With early and accurate detection of seizures, doctors can gain valuable time to administer medications and other such anti-seizure countermeasures to help reduce the damaging effects of this crippling disorder. The time-varying dynamics and high inter-individual variability make early prediction of a seizure state a challenging task. Many studies have shown that EEG signals do have valuable information that, if correctly analyzed, could help in the prediction of seizures in epileptic patients before their occurrence. Several mathematical transforms have been analyzed for its correlation with seizure onset prediction and a series of experiments were done to certify their strengths. New algorithms are presented to help clarify, monitor, and cross-validate the classification of EEG signals to predict the ictal (i.e. seizure) states, specifically the preictal, interictal, and postictal states in the brain. These new methods show promising results in detecting the presence of a preictal phase prior to the ictal state.

The mapping of sea ice is an important task for understanding global climate and for safe shipping. Currently, sea ice maps are created by human analysts with the help of remote sensing imagery, including synthetic aperture radar (SAR) imagery. While the maps are generally correct, they can be somewhat subjective and do not have pixel-level resolution due to the time consuming nature of manual segmentation. Therefore, automated sea ice mapping algorithms such as the multivariate iterative region growing with semantics (MIRGS) sea ice image segmentation algorithm are needed. MIRGS was designed to work with one-channel single-polarization SAR imagery from the RADARSAT-1 satellite. The launch of RADARSAT-2 has made available two-channel dual-polarization SAR imagery for the purposes of sea ice mapping. Dual-polarization imagery provides more information for distinguishing ice types, and one of the channels is less sensitive to changes in the backscatter caused by the SAR incidence angle parameter. In the past, this change in backscatter due to the incidence angle was a key limitation that prevented automatic segmentation of full SAR scenes. This thesis investigates techniques to make use of the dual-polarization data in MIRGS. An evaluation of MIRGS with RADARSAT-2 data was performed and showed that some detail was lost and that the incidence angle caused errors in segmentation. Several data fusion schemes were investigated to determine if they can improve performance. Gradient generation methods designed to take advantage of dual-polarization data, feature space fusion using linear and non-linear transforms as well as image fusion methods based on wavelet combination rules were implemented and tested. Tuning of the MIRGS parameters was performed to find the best set of parameters for segmentation of dual-polarization data. Results show that the standard MIRGS algorithm with default parameters provides the highest accuracy, so no changes are necessary for dual-polarization data. A hierarchical segmentation scheme that segments the dual-polarization channels separately was implemented to overcome the incidence angle errors. The technique is effective but requires more user input than the standard MIRGS algorithm.

A practical and important class of scientific images are the 2D/3D images obtained from porous materials such as concretes, bone, active carbon, and glass. These materials constitute an important class of heterogeneous media possessing complicated microstructure that is difficult to describe qualitatively. However, they are not totally random and there is a mixture of organization and randomness that makes them difficult to characterize and study. In order to study different properties of porous materials, 2D/3D high resolution samples are required. But obtaining high resolution samples usually requires cutting, polishing and exposure to air, all of which affect the properties of the sample. Moreover, 3D samples obtained by Magnetic Resonance Imaging (MRI) are very low resolution and noisy. Therefore, artificial samples of porous media are required to be generated through a porous media reconstruction process. The recent contributions in the reconstruction task are either only based on a prior model, learned from statistical features of real high resolution training data, and generating samples from that model, or based on a prior model and the measurements. The main objective of this thesis is to some up with a statistical data fusion framework by which different images of porous materials at different resolutions and modalities are combined in order to generate artificial samples of porous media with enhanced resolution. The current super-resolution, multi-resolution and registration methods in image processing fail to provide a general framework for the porous media reconstruction purpose since they are usually based on finding an estimate rather than a typical sample, and also based on having the images from the same scene -- the case which is not true for porous media images. The statistical fusion approach that we propose here is based on a Bayesian framework by which a prior model learned from high resolution samples are combined with a measurement model defined based on the low resolution, coarse-scale information, to come up with a posterior model. We define a measurement model, in the non-hierachical and hierarchical image modeling framework, which describes how the low resolution information is asserted in the posterior model. Then, we propose a posterior sampling approach by which 2D posterior samples of porous media are generated from the posterior model. A more general framework that we propose here is asserting other constraints rather than the measurement in the model and then propose a constrained sampling strategy based on simulated annealing to generate artificial samples.

The objective of this work is to develop mathematical simulation models for predicting the transient behaviour of strings and beams subjected to impacts. The developed models are applied to study the dynamics of the piano and the sitar. For simulating rigid point impacts on continuous systems, a new method is proposed based on the unit impulse response. The developed method allows one to relate modal velocities before and after impact, without requiring the integration of the system equations of motion during impact. The proposed method has been used to model the impact of a pinned-pinned beam with a rigid obstacle. Numerical simulations are presented to illustrate the inability of the collocation-based coefficient of restitution method to predict an accurate and energy-consistent response. The results using the unit-impulse-based coefficient of restitution method are also compared to those obtained with a penalty approach,with good agreement. A new moving boundary formulation is presented to simulate wrapping contacts in continuous systems impacting rigid distributed obstacles. The free vibration response of an ideal string impacting a distributed parabolic obstacle located at its boundary is analyzed to understand and simulate a sitar string. The portion of the string in contact with the obstacle is governed by a different partial differential equation (PDE) from the free portion represented by the classical string equation. These two PDEs and corresponding boundary conditions, along with the transversality condition that governs the dynamics of the moving boundary, are obtained using Hamilton's principle. A Galerkin approximation is used to convert them into a system of nonlinear ordinary differential equations, with time-dependent mode-shapes as basis functions. The advantages and disadvantages of the proposed method are discussed in comparison to the penalty approach for simulating wrapping contacts. Finally, the model is used to investigate the mechanism behind the generation of the buzzing tone in a sitar. An alternate formulation using the penalty approach is also proposed, and the results are contrasted with those obtained using the moving boundary approach. A model for studying the interaction between a flexible beam and a string at a point including friction has also been developed. This model is used to study the interaction between a piano hammer and the string. A realistic model of the piano hammer-string interaction must treat both the action mechanism and the string. An elastic stiff string model is integrated with a dynamic model of a compliant piano action mechanism with a flexible hammer shank. Simulations have been used to compare the mechanism response for impact on an elastic string and a rigid stop. Hammer head scuffing along the string, as well as length of time in contact, were found to increase where an elastic string was used, while hammer shank vibration amplitude and peak contact force decreased. Introducing hammer-string friction decreases the duration of contact and reduces the extent of scuffing. Finally, significant differences in hammer and string motion were predicted for a highly flexible hammer shank. Initial contact time and location, length of contact period, peak contact force, hammer vibration amplitude, scuffing extent, and string spectral content were all influenced.