In this paper, we study the effects of finite size resources in point-to-point energy harvesting wireless networks. We consider a transmitter and a receiver with finite size buffers and batteries, both powered by energy harvesters. We use a discrete-time analysis technique to study total packet loss at the transmitter and receiver when both are constrained by finite resources. We show how resources at the transmitter can be traded off with resources at the receiver to achieve best performance. We characterize the benefits of feedback and show that feedback is beneficial at smaller battery and buffer sizes, and when the energy harvested is highly varying.

In this paper, we obtain the capacity region for the AWGN Multiple Access Channel (MAC) with transmitters equipped with energy harvesters. We next obtain the sum-of-rates capacity for the flat fading Multiple Access Channel with energy harvesting transmitters. We then develop low delay, adaptive transmission strategies that surprisingly achieve near optimal performance under finite battery resource constrained settings. We study the effect of asymmetries in the energy harvesting and characterize the achievable performance.

We derive a lower bound on the diversity-multiplexing tradeoff for multiple antenna multiple access channels using temporal power control (PC). We quantify the substantial improvements in diversity gains with finite rate feedback over schemes that use no feedback, even when users have asymmetric diversity and rate requirements. For small multiplexing gains, all users in the MAC achieve their single user performance as if each receives B distinct bits of feedback. We also show that for a SISO MAC with finite rate feedback, the diversity orders with PC and with opportunistic user selection are the same at all multiplexing gains.

In this paper, we study the relationship between queuing delay for a random packet arrival process and physical layer parameters in a multiple access channel with multiple antennas at the input and output. Our main contribution is the derivation of a simple, analytical approximation for the average delay that clearly indicates the effect of number of transmit and receive antennas, transmission power, the packet arrival rate and the desired reliability. Comparison with numerical analysis indicates that the proposed analytical approximation of the delay is accurate for medium and large SNRs.

We consider a multiple access channel (MAC) with multiple antennas, where each user has different diversity and multiplexing gain requirement. For this configuration, we characterize the fundamental tradeoff region for each user. Specifically, we compute the maximum achievable diversity gain for a user in a MAC, given an achievable point in the multiplexing gain region.

In this paper, we investigate the optimal tradeoff between source and channel coding for channels with bit or packet erasure. Upper and lower bounds on the optimal channel coding rate are computed to achieve minimal end-to-end distortion. The bounds are calculated based on a combination of sphere packing, straight line and expurgated error exponents and also high rate vector quantization theory. By modeling a packet erasure channel in terms of an equivalent bit erasure channel, we obtain bounds on the packet size for a specified limit on the distortion

This research examines a variety of approaches for using two-dimensional orthogonal polynomials for the recognition of handwritten Arabic numerals. It also makes use of parametric and non-parametric statistical and neural network classifiers. Polynomials, including Legendre, Zernike, and pseudo-Zernike, are used to generate moment-based features which are invariant to location, size, and (optionally) rotation. An efficient method for computing the moments via geometric moments is presented. A side effect of this method also yields scale invariance. A new approach to location invariance using a minimum bounding circle is presented, and a detailed analysis of the rotational properties of the moments is given. Data partitioning tests are performed to evaluate the various feature types and classifiers. For rotational invariant character recognition, the highest percentage of correctly classified characters was 91.7%, and for non-rotational invariant recognition it was 97.6%. This compares with a previous effort, using the same data and test conditions, of 94.8%. The techniques developed here should also be applicable to other areas of shape recognition.

We present a new method for channel equalization using fuzzy logic. The membership functions are derived from the training data set, and a method to estimate the delay of the communication channel is proposed. The performance of the fuzzy equalizer is compared with that of a transversal filter equalizer. It is shown using simulations that the transversal filter requires a much larger training set to achieve the same error rate. It is also shown, using simulations, that the fuzzy equalizer performs better in the presence of channel nonlinearities.>

The advent of high-performance digital signal processors has spurred the development of sophisticated algorithms for adaptive filtering which are finding application in many diverse areas. Adaptive echo cancelers, which are employed to preserve voice quality in telecommunication systems, demand fast convergence to minimize the subjective effects of uncancelled echo at the beginning of a conversation. The LMS adaptive algorithm, which is simple to implement, suffers from slow convergence. The RLS algorithm, which provides faster convergence, is prohibitively expensive to implement. It is shown that application of the conjugate gradient adaptation algorithm for echo cancellation results in fast convergence with reasonable complexity.>

The paper describes a vision system for automatic inspection of the connecting part of the wire bond of an IC where the wire connects to the bond pad on the chip. It considers a popular type of such bonds known as 'ball bond'. Using two-dimensional images taken from the top of the IC wafer, the system determines several geometric measures which are important in determining the quality of the bond. These measures include the boundary, length of major and minor axes of the best fitting ellipse and the center. The process utilizes automatic thresholding, morphological operations and geometric moments of the image. Success of the method is demonstrated through experimental studies on actual bonds.>

In this paper, the problem of restoration of multiple misregistered, blurred and noisy images is considered. Motivation for this problem comes from the larger problem of construction of a high resolution restored image from the observed set of low-resolution images. The global approach is to decompose the problem into a sequence of sub-problems, namely registration, restoration and interpolation. The restoration problem is addressed here. For simultaneous restoration of the observed images, a parametric stochastic model for the image correlations is derived and used to obtain a linear minimum mean square error (LMMSE) estimate of the original image at each sensor. The resulting restored low-resolution images are integrated to construct a single restored high resolution image. Simulation results are presented to illustrate the performance of the proposed restoration scheme.

An algorithm is presented to recognize and locate partially distorted 2D shapes without regard to their orientation, location, and size. The algorithm first calculates the curvature function from the digitized image of an object. The points of local maxima and minima extracted from the smooth curvature are used as control points to segment the boundary and to guide the boundary-matching procedure. The boundary-matching procedure considers two shapes at a time, one shape from the template databank, and the other from the object being classified. The procedure tries to match the control points in the unknown shape to those of a shape from the template databank, and estimates the translation, rotation, and scaling factors to be used to normalize the boundary of the unknown shape. The chamfer 3/4 distance transformation and a partial distance measurement scheme constitute the final step in measuring the similarity between the two shapes. The unknown shape is assigned to the class corresponding to the minimum distance. The algorithm has been successfully tested on partial shapes using two sets of data, one with sharp corners and the other with curve segments. This algorithm not only is computationally simple, but also works reasonably well in the presence of a moderate amount of noise.>

A compound Gauss-Markov random field (CGMRF) that models nonstationarity in images is proposed for the segmentation and restoration of blurred and noisy images. At the top level of the CGMRF, a label process, which segments the image into K regions, is modeled by a Gibbs random field (GRF). At the bottom level, pixel intensities in each region are modeled by a stationary, noncausal GMRF. maximum a posteriori (MAP) estimates of segmented and restored images are obtained by maximizing their joint a posteriori distribution, using model parameters identified from the noisy image. A stochastic relaxation method is used for optimization. For faster convergence, deterministic relaxation is implemented. Experimental results on segmenting and restoring a noisy image are presented.>

It is well known that constrained recursion techniques can be used to restore images degraded by convolutional filters. The recursion which is commonly used was designed to work on a single noise-free image frame and convergence conditions were derived via the contraction mapping theorem. However these conditions do not guarantee convergence when the degrading filter has zeros in its transfer function. The authors consider the case where the imaging system gathers a sequence of noisy images of a static scene. The problem formulation uses a state space approach to provide easily verifiable conditions on the degrading filter which guarantee that the scene can be recovered from the image sequence in the presence of noise. It is shown that even transfer functions which have zeros are allowed. A recursive filter is developed to construct the estimate of the scene from the image sequence and experimental results are given.>

A technique for classifying closed planar shapes is described in which a shape is characterized by an ordered sequence that represents the Euclidean distance between the centroid and all contour pixels of the shape. Shapes belonging to the same class have similar sequences, hence a procedure for classifying shapes is based on the degree of similarity between these sequences. In order to determine the similarity between sequences, a non-linear alignment process is developed to find the best correspondence between the sequences. Optimum alignment is obtained by expanding segments of the sequences to minimize a dissimilarity function between the sequences. Normalization with respect to scaling and rotation is described and an example illustrating the use of dynamic alignment for the classification of noisy shapes is presented.

Source encoding of images for bandwidth compression has become attractive in recent years because of decreasing hardware costs. By combining the source encoding approach with transform coding techniques, it is possible to obtain good image quality at low data rates. The general aspects of such a system are presented. The design of the quantizer for transform coefficients, which is the major source of error associated with the compression process, is considered using a visual fidelity criterion and subject to the constraint that the entropy of the quantizer be a prespecified quantity. A visually weighted suboptimal quantization scheme is developed to take into account the relative importance of different transform coefficients to the human visual system.

This paper presents a recursive image enhancement algorithm based on two-demensional (2-D) sequential suboptimal estimation using a spatially varying Markov model. An algorithm for identifying the parameters of a class of monochromatic images modeled by a nonsymmetric half-plane (NSHP) model is derived and used in conjunction with a simple estimation scheme to obtain a locally suboptimal enhancer. The enhancement algorithm is applied to several images to test its validity.

The problem of estimating the spectrum of a complex stochastic process observed in additive white noise is considered. The problem is reduced to one of parameter identification by assuming an autoregressive model for the process. An algorithm for estimating these parameters is derived. Examples are presented to illustrate the use of the algorithm.

The use of two-dimensional interpolative models for recursive enhancement of noisy images is considered. The two-dimensional data is decorrelated either row-wise or column-wise to obtain an equivalent one-dimensional interpolative model. An adaptive identification scheme to identify the parameters of the model is presented which are then used in a simple estimation scheme for recovering the image data. A modification of the FFT algorithm as well as the DCT algorithm are used in decorrelating the data. The performance of these transforms for the present application is compared in terms of the number of computations required and the mean-square error between the enhanced image and the original noise-free image. Several examples are presented to illustrate the applicability of the algorithm.

A sequential estimation algorithm for the enhancement of noisy images is presented. A set of parameters obtained by a transformation of an autoregressive (AR) model representing the image process, is used for computational advantages. The method presented has an added advantage that it requires no a priori knowledge of the statistics of the noise-free data. Several examples are presented to illustrate the utility of the algorithm.

The problem of tracking a target in a multitarget environment when the observations are received over a fading channel is considered. The optimal Bayes solution to the tracking problem in such cases involves growing memory and hence is not feasible. A particularly effective suboptimal scheme uses a probabilistic judgment at each stage of the observations to overcome this problem. This concise paper presents an evaluation of the scheme in terms of mean-square error performance when the observations are received over fading channels.

This paper considers binary pattern recognition of a non-Gaussian pattern in Gaussian noise using supervised learning. The scheme is both structure and parameter adaptive. To facilitate a feasible solution, certain judicious approximations are used. Two examples are presented to demonstrate the learning capability of the proposed algorithms.

A new formulation of the problem of structure and parameter adaptive pattern recognition with supervised learning yielding significant computational advantages is presented.

This note applies the general likelihood ratio for the minimum-mean-square-error estimation of the signal when the presence of the signal at the receiver is uncertain during the entire observation interval.