We consider the problem of detecting small and manoeuvring objects with staring array radars. Coherent processing and long-time integration are key to addressing the undesirably low signal-to-noise/background conditions in this scenario and are complicated by the object manoeuvres. We propose a Bayesian solution that builds upon a Bernoulli state space model equipped with the likelihood of the radar data cubes through the radar ambiguity function. Likelihood evaluation in this model corresponds to coherent long-time integration. The proposed processing scheme consists of Bernoulli filtering within expectation maximisation iterations that aims at approximately finding complex reflection coefficients. We demonstrate the efficacy of our approach in a simulation example.

In this article, we introduce a frequency-based waveform design to maximize binary target classification. The optimization problems can be solved via optimal solvers available in the literature. The presented formulation is applicable to the classification scenarios where the extended target frequency responses (TFRs) are complex, random, and normally distributed with unequal mean vectors. However, their covariance matrices are either identical or different. For the former, i.e., identical covariance matrices, the optimization problem consists of a cost function that has been obtained by deriving a closed-form expression of the probability of misclassification. For the latter, i.e., different covariance matrices, we present an optimization problem where the cost function has been obtained via the Fisher separation function. For this, we also present a closed-form expression for an approximate solution to the optimization problem. We show that the proposed solution achieves performance levels comparable to the exact solution of the optimization problem obtained via state-of-the-art optimization solvers while incurring low computational complexity. We expand on the two main scenarios by introducing clutter (i.e., signal-dependent interference) in the signal model and studying possible closed-form designs in extreme waveform energy levels. Simulations are conducted using synthetically generated data in addition to the civilian vehicle data from the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset, in order to validate the proposed method.

In this article, we propose new waveform design procedures and classification schemes to improve target classification performance nonadaptively in radar systems. The new designs and schemes are all inspired by 2-class and multiclass Fisher discriminant analysis. The proposed system does not require as much computational capability as adaptive waveform design systems while also overcoming: 1) angular uncertainty in classifying high fidelity targets and 2) drops in performance experienced by nonadaptive systems when classification is extended to more than two targets. The waveform design procedure is based on an optimization problem to find the waveform that maximizes the objective function inspired by Fisher analysis under constant energy constraint. We also derive two closed-form solutions for the optimization problems under certain conditions for the 2-class and multiclass cases. All the methods are tested using synthetic and real data to show the performance of the proposed methods against the average Mahalanobis distance (AMD) nonadaptive waveform design and classifier.

In this letter, we propose graph signal processing-based imaging for synthetic aperture radar (SAR). Our method provides improved denoising and resolution enhancing capabilities, along with a reduction in computational complexity, by exploiting the concept of extended neighborhood in SAR images. We present a modified version of a fused least absolute shrinkage and selection operator (LASSO) to cater for graph structure of the SAR image. It can also accommodate the compressed sensing framework. We solve the optimization problem via the alternating direction method of multipliers. Experimental results on a backhoe target corroborate the validity of our proposed method.

A frequency-domain system identification-based multiple-input multiple-output (FDSI-MIMO)-SAR algorithm using multiple phase center multiple azimuth beams is proposed to obtain high-resolution wide-swath (HRWS) imaging. Frequency-division multiple access is used in such a way that each transmitter emits a conventional linear frequency-modulated (LFM) waveform modulated by a different carrier frequency and the obtained range resolution corresponds to the total transmitted bandwidth. In this article, an MIMO-SAR problem is modeled, using the principle of displaced phase center, as multiple but separate multiple-input single-output (MISO) system identification problems. The channel impulse responses of the individual MISO problems are identified in the range dimension using an FDSI-based estimation algorithm in such a way that the estimated range profile is free of interrange cell interference. In addition, the proposed algorithm does not require separating the subband waveforms at the receiver as they are processed jointly without a need to add guard bands between the adjacent subbands, which would allow utilizing the available bandwidth to the maximum efficiency. The method of synthesizing a wide transmit antenna beam from a narrow antenna beam applied in single-input multiple-output (SIMO) SAR is extended to MIMO SAR to remove the sidelobes effects of the receive beams and hence leading to the desired signal-to-noise ratio. A pulse repetition frequency lower than the Doppler bandwidth is used to obtain wide swath without experiencing aliasing in the Doppler spectrum. Finally, both simulated and constructed raw data are used to validate the effectiveness of the proposed algorithm.

In this paper, an optimal radar waveform-design scheme is proposed, using frequency snapshots (i.e., frequency-domain processing), based on an objective function derived directly from a binary classification/identification criterion constrained by waveform-energy. We consider an extended-target model where the targets frequency response is assumed to be complex Gaussian. High- and low-energy solutions for the optimal waveform are explored and closed-form expressions in both the scenarios are derived.

This paper proposes a method for minimum mean squared error (MMSE) adaptive waveform design (AWD) in multiple-input-multiple-output (MIMO) active sensing systems which are used to track moving targets. The method proposed herein prompts two computational improvements compared to a related method for static targets. Consideration of moving targets also introduces the possibility of `model mismatch' between the actual motion of the targets, and the model available to the MMSE AWD system. Results show that the proposed method leads to an improvement in mean squared error performance of up to 29% compared to the non-adaptive case.

In this work, we consider the problem of synchronising separately located transmitters and a staring array receiver that also has a local transmitter. The acknowledged benefits of using separate transmitters in active sensing are often undermined by the difficulty in accurate synchronisation of the receiver and the transmitters. In this work, we propose a solution that is based on measurements from non-cooperative objects in the illuminated region. We formulate the problem as parameter estimation in a state space model with individual transmitter channel data cubes as measurements. For maximum likelihood estimation, we use an expectation maximisation type iterative bound optimisation using distributions found by track-before-detect together with explicit formulae derived here for the related score function for chirp waveforms. We demonstrate that the proposed approach is capable of achieving very high accuracy with errors on the order of small fractions of the pulse width thereby enabling coherent processing in bi-static channels.

Motivated by object tracking applications with networked sensors, we consider multi sensor state space models. Estimation of latent parameters in these models requires centralisation because the parameter likelihood depend on the measurement histories of all of the sensors. Consequently, joint processing of multiple histories pose difficulties in scaling with the number of sensors. We propose an approximation with a node-wise separable structure thereby removing the need for centralisation in likelihood computations. When leveraged with Markov random field models and message passing algorithms for inference, these likelihoods facilitate decentralised estimation in tracking networks as well as scalable computation schemes in centralised settings. We establish the connection between the approximation quality of the proposed separable likelihoods and the accuracy of state estimation based on individual sensor histories. We demonstrate this approach in a sensor network self-localisation example.

First results of repeat pass SAR Interferometry (InSAR) from Selex ES's compact, experimental airborne synthetic aperture radar (SAR) system are described for civilian environmental monitoring. The system is a low cost, lightweight (<;10kg), X-band, Active Electronically Scanned Array (AESA) SAR platform, which provides greater opportunities of access to InSAR data as it can be placed on helicopters, fixed wing aircrafts and UAVs. Results of high resolution (<; 2m) digital elevation models (DEMs) produced are shown for two sites of environmental interest. The first DEMs produced compare well to those from other datasets.

As multiple-input multiple-output (MIMO) radar gains popularity, more efficient and better-performing detection algorithms are developed to exploit the benefits of having more transmitters and receivers. Many of these algorithms are based on the assumption that the multiple waveforms used for target scanning are orthogonal to each other in fast time. It has been shown that this assumption can limit the practical detector performance due to the reduction of the area that is clear of sidelobes in the MIMO radar ambiguity function. In this work it is shown that using the same waveform with a different carrier frequency and/or delay across different transmitters ensures relative waveform orthogonality while alleviating the negative effects on the ambiguity function. This is demonstrated in a practical scenario where the probing waveforms consist of Gaussian pulse trains (GPTs) separated in frequency. An approximate theoretical model of the ambiguity is proposed and it is shown that the effects of cross-ambiguity in the MIMO system are negligible compared to the waveform autoambiguities.

The data collected by sensor nodes over a geographical region is contaminated with Gaussian and impulsive noise. The conventional gradient-based distributed adaptive estimation algorithms exhibit good performance in the presence of Gaussian noise but perform poorly in impulsive noise environments. Therefore, the objective of this article is to propose a robust distributed adaptive algorithm that alleviates the effect of impulsive noise. An error saturation nonlinearity-based robust distributed strategy is proposed in an incremental cooperative network to estimate the desired parameters in impulsive noise. The steady-state analysis of the proposed error saturation nonlinearity incremental least mean squares (SNILMS) algorithm is carried out by employing the spatial-temporal energy conservation principle. Both theoretical and simulation results show that the presence of the error nonlinearity has made the proposed SNILMS algorithm robust to impulsive noise.

This paper describes and supplements a recent re-examination of linear time-frequency decomposition wherein the principle of stationary phase is applied to the synthesis integral. An inherent part of this time-frequency stationary phase approximation (TFSPA) is a test for the stationary phase condition itself. After outlining the development of the TFSPA, the main contribution of this paper is an analysis of the test from the perspective of classical detection theory. This leads to closed form approximations that; (i) quantify performance in terms of false alarm and detection probabilities; (ii) enable the development of improved tests.

Multiple input multiple output (MIMO) radars have many advantages over their phased-array counterparts: improved spatial resolution; better parametric identifiably and greater flexibility to acheive the desired transmit beampattern. The desired transmit beampatterns using MIMO-radar requires the waveforms to have arbitrary auto- and cross-correlations. To design such waveforms, generally a waveform covariance matrix, R, is synthesised first then the actual waveforms are designed. Synthesis of the covariance matrix, R, is a constrained optimisation problem, which requires R to be positive semidefinite and all of its diagonal elements to be equal. To simplify the first constraint the covariance matrix is synthesised indirectly from its square-root matrix U, while for the second constraint the elements of the m-th column of U are parameterised using the coordinates of the m-hypersphere. This implicitly fulfils both of the constraints and enables us to write the cost-function in closed form. Then the cost-function is optimised using a simple particle-swarm-optimisation (PSO) technique, which requires only the cost-function and can optimise any choice of norm cost-function.

Calibration of array shape error is a key issue for most existing source localisation algorithms. In this study, the far-field self-calibration and near-field pilot-calibration are carried out using unconditional maximum likelihood (UML) estimator whose objective function is optimised by particle swarm optimisation (PSO). A new technique, decaying diagonal loading (DDL), is proposed to enhance the performance of PSO at high signal-to-noise ratio (SNR) by dynamically lowering it, based on the counter-intuitive observation that the global optimum of the UML objective function is more prominent at lower SNR. Numerical simulations demonstrate that the UML estimator optimised by PSO with DDL is robust to large shape errors, optimally accurate and free of the initialisation problem. In addition, the DDL technique can be coupled with different global optimisation algorithms for performance enhancement. Mathematical analysis indicates that the DDL is applicable to any array processing problem where the UML estimator is employed.

In the important domain of array shape calibration, the near-field case poses a challenging problem due to the array response complexity induced by the range effect. In this paper, near-field calibration is carried out using an unconditional maximum likelihood (UML) estimator. Its objective function is optimized by the particle swarm optimization (PSO) algorithm. A new technique, decaying diagonal loading (DDL) is proposed to enhance the performance of PSO at high signal-to-noise ratio (SNR) by dynamically lowering it, based on the counter-intuitive observation that the global optimum of the UML objective function is more prominent at lower SNR. UML estimator offers Cramér-Rao bound (CRB)-attaining accuracy. The direct optimization by PSO without approximation makes the estimator applicable to the entire near-field. In addition, PSO is free of the initialization problem from which the local optimization algorithms suffer. Numerical simulations demonstrate the CRB-attaining results at SNR as high as 60 dB.

A method for generating constant envelope (CE) waveforms to realise a given covariance matrix for a closely spaced MIMO radar system is proposed. In contrast to available algorithms, the technique provides closed form solutions for finding the required waveforms and suggests that waveforms can be chosen from finite alphabets such as binary-phase shift keying (BPSK) and quadrature-phase shift keying (QPSK). Gaussian random-variables (RV's) are mapped onto CE non-Gaussian RV's using memoryless non-linear functions. The relationship between the correlation of Gaussian RV's at the input to the nonlinear functions and non-Gaussian RV's at their output is established. Simulation results are presented to demonstrate the effectiveness of the methodology.

We consider the use of arbitrary 3D arrays for direction-of-arrival (DOA) estimation, with the commonly used Cramér-Rao Bound (CRB) as the main performance measure. The contribution of this paper is a set of techniques for analyzing the benefits of sensor relocation over uncertain terrain, more specifically where the random elevation is unknown outside of current locations. Suitable arrays for data collection are defined and generated. Measures to assess the effects of planar coordinate improvements and sensor elevations are defined. Descriptions of suitable visualization formats to interpret these measurements are given, and some illustrative results support the utility of the proposed techniques.

This paper introduces two new distributed learning algorithms : Incremental Particle Swarm Optimization (IPSO) and Diffusion Particle Swarm Optimization (DPSO). These algorithms are applied for distributed identification of nonlinear processes using cooperation among adaptive nodes. Identification of four standard nonlinear plants have been carried out through simulation to assess the performance of these algorithms. The results indicate better or identical identification performance offered by the proposed distributed algorithms compared to that offered by the conventional PSO based algorithm. The improvement is observed in terms of CPU time, accuracy in response matching and speed of convergence.

We present an algorithm that minimizes asymptotically a sequence of non-negative convex functions over diffusion networks. To account for possible node failures, position changes, and/or reachability problems (because of moving obstacles, jammers, etc), the algorithm can cope with dynamic networks and cost functions, a desirable feature for online algorithms where information arrives sequentially. Many projection-based algorithms can be straightforwardly extended to diffusion networks with the proposed scheme. We use the acoustic source localization problem in sensor networks as an example of a possible application.

We attempt to quantify the possible gains that can be achieved by examining a rate-distortion competition between a conventional and a compressive sampling solution to data rate reduction. Simple approximate expression are developed for the minimum bit rate required to obtain the best achievable average performance from the compressive sensing system and the performance that would be achieved if that rate requirement was met. An example of a signal that contains a small number of phasors in Gaussian white noise is used to validate these results and to compare the degradation in performance of the 2 systems when lower bit rates than those required by the theory are employed.

For bistatic ground moving target indication radar, the clutter Doppler frequency depends on range for all array geometries. This range dependency leads to problems in clutter suppression through STAP techniques. In this paper, we propose a new approach of applying non-linear prediction theory to address the range dependency problem in bistatic airborne radar systems. This technique uses a non-linear function to obtain an estimate of the range-dependent inverse covariance matrix. Simulation results suggest a non-linear fit for the model (nonlinear relationship between the inverse covariance matrices) and show an improvement in processor performance as compared to conventional STAP methods.

Traditional space time adaptive processors for radar target detection require a training data set which is usually drawn from adjacent range gates. Clutter heterogeneity, however, can severely limit the available training sample support and consequently degrade the detection performance. The SDS algorithms, on the other hand, overcome this problem by operating solely on the test data without recourse to training data. In this paper we evaluate both of these approaches, in particular the AMF and MLED, using the MCARM data set. We illustrate the performance degradation of the AMF that results from the clutter heterogeneity and the corresponding advantage of the MLED. We also show that a calibration step of the spatial steering vectors results in significant performance improvement of all of the algorithms considered here.

A maximum-likelihood channel estimator for the orthogonal frequency division multiplexing communication environments, in the presence of interference is discussed here. In a training-based scenario, the channel is estimated based on pilots that precede the transmission of the information. To reduce the number of estimation parameters, the channel is estimated iteratively in time-domain. Since interference from other users provide no useful information, parameters of the interference are neither estimated nor the effect of the interference neglected, instead interference along with Gaussian noise is perceived as non-Gaussian noise process. The algorithm assumes no a priori knowledge about the interfering channel and signal at the receiver, further no assumption on the statistical properties of the interferer is assumed, which makes this algorithm robust. The estimated channel information along with the estimated distribution are then utilised to equalise the subsequent data blocks. The strength of the algorithm is in its robustness to both synchronous and asynchronous interference, which is confirmed by the simulation results for both flat and multipath fading channels in presence of synchronous and asynchronous interference.

In this paper, a new fast adaptive blind channel estimation method is proposed using the subspace information from the correlation matrix. The algorithm is fully adaptive in the sense that both the subspace information and the optimization which leads to the channel estimation are computed adaptively. It is based on the recently proposed YAST subspace tracker which has been shown to outperform other methods both in terms of speed of convergence and computational complexity. A discussion on the convergence properties of the proposed algorithm is presented. We also propose a hybrid method which makes use of the YAST subspace tracker for initial fast convergence and the subspace information is then updated using the numerically stable OPAST subspace tracker.

It is well established that the quality of the channel estimate plays a crucial role in the performance of a rake receiver. This paper addresses the problem of optimising the channel estimate for a wideband DS–CDMA rake receiver equipped with a sliding–window adaptive channel estimator. The mean squared error of the channel estimate is analytically extracted providing the means to optimise the window size in a minimum–MSE sense. Unfortunately, the BER at the output of the rake receiver is only partially analytically tractable, but minimum–BER optimisation of the window size is experimentally shown to outperform the MMSE criterion.

Principal component analysis (PCA) has been used in many applications ranging from social science to space science, for the purpose of data compression and feature extraction. Usage of PCA for synthetic aperture radar (SAR) image classification, though widely reported by remote-sensing researchers, has not been exploited much by automatic target recognition (ATR) community. In the present paper, PCA has been used in SAR-ATR using the MSTAR data base, and comparison has been made with the conventional conditional Gaussian model based Bayesian classifier (M.D. DeVore and J.A. O'Sullivan, 2002). The results have been compared based on percentage of correct classification, receiver operating characteristics (ROC), and performance with limited amount of training data. By all standards of comparison, the PCA based classifier was observed to outperform the conditional Gaussian model based Bayesian classifier (CGBC) or at the worst it performs at par. And given the computational and algorithmic simplicity of PCA based classifier, the new algorithm was concluded to be a highly prospective candidate for real time ATR systems

In this letter, we study issues associated with applying least-squares estimation to predict the inverse covariance matrix in bistatic airborne radar systems. For the bistatic ground moving target indication radar, the clutter Doppler frequency depends on the range for all array geometries. This range dependency leads to problems in clutter suppression through space-time adaptive processing (STAP) techniques. This paper proposes a new method of obtaining an estimate of the inverse covariance matrix using linear prediction techniques. Simulation results show a significant improvement in processor performance as compared to conventional STAP methods.

A perceptually motivated method is proposed for solving the permutation ambiguity of frequency-domain independent component analysis when the mixing environment is noisy and reverberant. In this method, perceptually irrelevant frequencies are removed from the speech spectrum using block based perceptual masking (simultaneous frequency masking) and then independent component analysis is applied. After source separation in frequency domain, a physical property of the mixing matrix, i.e., the coherency in adjacent frequencies, is utilized to solve the permutation ambiguity. From the simulation results it appears that the perceptual masking avoids the permutation problem.

The use of antenna arrays can help combat cochannel interference (CCI) in wireless cellular systems. In this paper, we consider an enhanced general packet radio service diversity receiver based on least squares spatio-temporal equalization and soft-decision decoding in the presence of decision feedback and/or asynchronous CCI. We compare known and novel estimators of the error mean and variance at the output of the deterministic space-time equalizer. The collected simulation data indicate that the estimation of the error mean and variance is critical to the performance of soft-in/hard-out Viterbi decoding in the presence of nonstationary input disturbance. Moreover, the use of short-term error statistics provides receiver performance gains of up to 15-20 dB in terms of signal-to-interference ratio, with respect to the use of burst statistics based on the training sequence midamble and tentative decisions on the payload symbols.

Independent component analysis (ICA) has shown success in the separation of sources in lots of applications. Almost all of them assume that a set of recorded signals is the result of a linear mixture of independent sources. Although ICA methods were firstly designed to apply only to free-noise signals, numerous methods have extended it to deal with additive noise, using only higher order statistics. However, in speckle environment signals the noise is multiplicative, so the applicability of ICA is seriously reduced. This paper proposes an ICA method for speckle signals, taking into account the multiplicative nature of the noise and improving the results obtained by standard ICA methods.

The demand for low-cost and low-power decoder chips has resulted in renewed interest in low-complexity decoding algorithms. In this paper a novel modification of the Max-Log-MAP algorithm is proposed for use in a turbo decoding process. This is achieved by scaling the a priori information by correction weights at each iteration, in order to maximize the exchange of mutual information between the component decoders. It is shown that the proposed technique results in a performance which approaches that of a turbo decoder using the optimum MAP algorithm, while maintaining the advantages of low complexity and insensitivity to input scaling inherent in the Max-Log-MAP algorithm. A second contribution of this paper is a method for off-line computation of the optimum weight values. The convergence behaviour of the proposed decoder is analysed via extrinsic information transfer (EXIT) charts.

We consider the problem of trellis equalization of the intersymbol interference channel in the presence of thermal noise and cochannel interference (CCI). Conventional maximum-likelihood sequence estimation (MLSE) and maximum a posteriori probability (MAP) trellis equalizers treat the sum of noise and interference as additive white Gaussian noise, while CCI is generally a colored non-Gaussian process. We propose a novel nonparametric approach based on the estimation of the probability density function of the noise-plus-interference. Given the availability of a limited volume of data, the density is estimated by kernel-smoothing techniques. The use of a whitening filter in the presence of temporally colored disturbance is also addressed. Simulation results are provided for the global system for mobile communications (GSM), showing a significant performance improvement with respect to the equalizer based on the Gaussian assumption. Major advantages of the proposed strategy are its intrinsic robustness and general applicability to those cases where accurate modeling of the interference is difficult or a model is not available.

The paper derives a stochastic-gradient minimum symbol-error-rate (MSER) algorithm, called the least symbol error rate (LSER), for training the linear equalizer and linear-combiner decision feedback equalizer (DFE) with M -PAM signalling. This LSER algorithm has some performance advantages, in terms of faster convergence rate and smaller steady-state symbol error rate (SER) misadjustment, over an existing simpler stochastic-gradient adaptive MSER algorithm called the approximate MSER (AMSER).

An adaptive minimum bit error rate (MBER) linear multiuser detector (MUD) is proposed for DS-CDMA systems. Based on the approach of kernel density estimation for approximating the bit error rate (BER) from training data, a least mean squares (LMS) style adaptive algorithm is developed for training linear MUDs. Computer simulation results show that this adaptive MBER linear MUD outperforms two existing LMS-style adaptive MBER algorithms.

Existing evidence for and against sea clutter being chaotic and nonlinearly predictable is briefly discussed. Despite the uncertainty surrounding the chaotic nature of sea clutter, and its nonlinear predictability, the purpose of this paper is to examine what the best design criterion is for a nonlinear predictor which is to be used to detect targets against clutter which is known to be chaotic: mean square error performance or capturing the chaotic clutter's underlying dynamics. Single pulse detection analysis using a Swerling I target and chaotic "clutter" is carried out using predictor-based detectors in an attempt to determine which criterion is most suitable. The predictor detectors are compared with standard detection strategies.

A detector based on multiple-hyperplane partitioning of the signal space is derived for realizing the optimal Bayesian decision feedback equaliser (DFE). It is known that the optimal Bayesian decision boundary separating any two neighbouring signal classes is asymptotically piecewise linear and consists of several hyperplanes, when the signal to noise ratio (SNR) tends to infinity. The proposed technique determines these hyperplanes and uses them to partition the observation space. The resulting detector can closely approximate the optimal Bayesian detector, at an advantage of considerably reduced decision complexity.

A decision-feedback equalizer scheme is derived based on multiple-hyperplane partitioning of signal space for detecting M-ary pulse amplitude modulation symbols transmitted through a noisy intersymbol interference channel. The proposed scheme is based on the fact that the optimal Bayesian decision boundary separating two neighboring signal classes is asymptotically piecewise linear and consists of several hyperplanes, when the signal-to-noise ratio tends to infinity. An algorithm is developed to determine these hyperplanes, which are then used to partition the observation signal space. The resulting detector can closely approximate the optimal Bayesian detector, at an advantage of considerably reduced detector complexity.

Detection algorithms for single user wireless communication using multiple antennas at both the transmitter and receiver in a Rayleigh (flat) fading environment are compared. The system includes N transmitting antennas, M receiving antennas (N/spl les/M) and repetition coding at the transmitter (delay diversity). The linear decorrelating (LD) and minimum mean squared error (MMSE) detectors are compared with their D-BLAST and V-BLAST versions using bit error ratio versus signal-to-noise ratio simulation. For BPSK, the MMSE detector and its BLAST versions perform best and yield almost indistinguishable BER curves. The LD detector performs worst. For fixed N, as M increases, all the BER curves converge. The effect of error propagation in the BLAST schemes is shown to be non-negligible.

We consider the reverse link of a cellular direct sequence CDMA system with the aim of increasing the system's capacity. The capacity of a conventional (matched filter) receiver can be increased by employing sectorized antenna arrays or multiuser detection techniques. Our results confirm that the potentially complex approach of beamforming followed by multiuser detection further increases capacity. To reduce complexity, bit detection techniques based on partitioning the users into spatial equivalence classes by beamforming and then employing standard multiuser detection techniques within each class are investigated. Subtractive interference cancellation yields the lowest bit error ratios and is the least computationally complex method of multiuser detection under consideration.

This work deals with investigating whether or not nonlinear predictor networks can be used to improve the performance of high resolution surveillance radars which are used to detect targets on, or near the sea surface. Prediction and detection results are presented for new sea clutter data sets.

This paper discusses the use of coherent demodulation for antenna array receivers employing direct sequence spread spectrum techniques and M-ary orthogonal modulation. Three different approaches are discussed. The first is a conventional receiver using training sequences and decision feedback for channel estimation. The other two techniques employ decisions from a simple noncoherent combining (O) receiver to circumvent the need for training sequences. The bit error ratio (BER) performance of coherent M-ary orthogonal modulation is derived. Some qualitative analysis is also presented for the algorithms, and simulation results are used to compare their performance in different scenarios.

We describe two algorithms for designing and training neural-network classifiers. The first, the linear programming slab algorithm (LPSA), is motivated by the problem of reconstructing digital signals corrupted by passage through a dispersive channel and by additive noise. It constructs a multilayer perceptron (MLP) to separate two disjoint sets by using linear programming methods to identify network parameters. The second, the perceptron learning slab algorithm (PLSA), avoids the computational costs of linear programming by using an error-correction approach to identify parameters. Both algorithms operate in highly constrained parameter spaces and are able to exploit symmetry in the classification problem. Using these algorithms, we develop a number of procedures for the adaptive equalization of a complex linear 4-quadrature amplitude modulation (QAM) channel, and compare their performance in a simulation study. Results are given for both stationary and time-varying channels, the latter based on the COST 207 GSM propagation model.

A nonlinear predictive approach has been employed in MPEG (Moving Picture Experts Group) video transmission in order to improve the rate control performance of the video encoder. A nonlinear prediction and quantisation technique has been applied to the video rate control which employs a transmission buffer for constant bit rate video transmission. A radial basis function (RBF) network has been adopted as a video rate estimator to predict the rate value of a picture in advance of encoding. The quantiser control surfaces based on nonlinear equations, which map both estimated and current buffer occupancies to a suitable quantisation step size, have also been used to achieve quicker responses to dramatic video rate variation. This scheme aims to adequately accommodate non-stationary video in the limited capacity of the buffer. Performance has been evaluated in comparison to the MPEG2 Test Model 5 (TM5) in terms of the buffer occupancy and picture quality.

The problem of active noise control in a linear duct is examined. Essentially, a nonlinear inverse to a nonminimum phase actuator is proposed. The nonlinear inverse exploits the non-Gaussian nature of some chaotic and stochastic noise sources. The architecture of the controller is derived using Bayesian estimation theory and is shown to be a combination of a linear adaptive network and a radial basis function (RBF) or Volterra series (VS) network. Because of the nonlinear nature of the controller, the filtered-x least means square (LMS) architecture cannot be used. Hence a modified active noise controller is proposed. Simulation results demonstrate the improvements in performance achievable with the combined linear and nonlinear controller.

This paper considers the performance of four channel identification techniques for code division multiple access (CDMA) antenna array receivers. These techniques are based on the assumption that the interference is spatially white: they provide a spatial "matched filter" solution. Perturbation formulae are presented for estimating the attainable signal to interference and noise ratios (SINR) for these techniques. Some simulation results are also presented to compare the convergence performance of these methods.

We present a method of modifying the structure of radial basis function (RBF) network to work with nonstationary series that exhibit homogeneous nonstationary behavior. In the original RBF network, the hidden node's function is to sense the trajectory of the time series and to respond when there is a strong correlation between the input pattern and the hidden node's center. This type of response, however, is highly sensitive to changes in the level and trend of the time series. To counter these effects, the hidden node's function is modified to one which detects and reacts to the gradient of the series. We call this new network the gradient RBF (GRBF) model. Single and multistep predictive performance for the Mackey-Glass chaotic time series were evaluated using the classical RBF and GRBF models. The simulation results for the series without and with a tine-varying mean confirm the superior performance of the GRBF predictor over the RBF predictor.

The paper investigates adaptive equalization of time-dispersive mobile radio fading channels and develops a robust high performance Bayesian decision feedback equalizer (DFE). The characteristics and implementation aspects of this Bayesian DFE are analyzed, and its performance is compared with those of the conventional symbol or fractional spaced DFE and the maximum likelihood sequence estimator (MLSE). In terms of computational complexity, the adaptive Bayesian DFE is slightly more complex than the conventional DFE but is much simpler than the adaptive MLSE. In terms of error rate in symbol detection, the adaptive Bayesian DFE outperforms the conventional DFE dramatically. Moreover, for severely fading multipath channels, the adaptive MLSE exhibits significant degradation from the theoretical optimal performance and becomes inferior to the adaptive Bayesian DFE.>

A multi-stage blind clustering algorithm is proposed for equalisation of multi-level quadrature amplitude modulation (M-QAM) channels. A hierarchical decomposition divides the task of equalising a high-order QAM channel into much simpler sub-tasks. Each sub-task can be accomplished fast and reliably using a blind clustering algorithm derived originally for 4-QAM signals. The constant modulus algorithm (CMA) is used as a benchmark to assess this multi-stage blind equaliser. It is demonstrated that the new blind algorithm achieves much faster convergence and is very robust when input symbols are not sufficiently white. This multi-stage clustering equaliser only requires slightly more computations than the CMA and, like the latter, its computational complexity does not increase as the levels of digital symbols increase.>

The orthogonal, least squares (OLS) algorithm is an efficient implementation of the forward regression procedure for subset model selection. The ability to find good subset parameters with only linear increase in computational complexity makes this method attractive for practical implementations. We examine the computation requirement of the OLS algorithm to reduce a model of K terms to a subset model of R terms when the number of training data available is N. We show that in the case where N/spl Gt/K, we can reduce the computation requirement by introducing an unitary transformation on the problem.>

A systematic approach to constructing a nonlinear adaptive filter is presented. The approach is based on a signal dependent orthonormal expansion implemented in two stages: (i) a signal independent standard orthonormal expansion; (ii) scaling using an estimate of the vector probability density function (pdf). Further it is demonstrated that the standard orthonormal function set can also provide an estimate of the pdf when used in conjunction with an inverse Fourier transform. The orthonormality has two implications for adaptive filtering: (i) model order reduction is trivial because the size of a coefficient in the final linear combiner is directly related to its contribution to the overall mean squared error; (ii) consistent, rapid convergence of stochastic gradient algorithms is assured. A typical nonlinear adaptive algorithm is presented.>

In conventional partially adaptive linearly constrained minimum variance (LCMV) beamformer design the approach has been to represent the noise subspace with some reduced set of vectors, typically the eigenvectors associated with the largest eigenvalues of the noise covariance matrix. This, whilst yielding good performance, will not give the optimum performance for a given partially adaptive dimension. The paper presents an alternative method for selecting the "best" degrees of freedom to be retained in a partially adaptive design. The iterative algorithm described selects those degrees of freedom which minimize the beamformer output mean square error. This approach leads to a sparse structure for the transformation matrix, which when implemented in a generalize sidelobe canceller (GSC) structure will reduce the computational load. This approach also allows a reduction in adaptive dimension as compared to the eigenvector based approach. An illustrative example demonstrates the effectiveness of this method.>

The application of a radial basis function network to digital communications channel equalization is examined. It is shown that the radial basis function network has an identical structure to the optimal Bayesian symbol-decision equalizer solution and, therefore, can be employed to implement the Bayesian equalizer. The training of a radial basis function network to realize the Bayesian equalization solution can be achieved efficiently using a simple and robust supervised clustering algorithm. During data transmission a decision-directed version of the clustering algorithm enables the radial basis function network to track a slowly time-varying environment. Moreover, the clustering scheme provides an automatic compensation for nonlinear channel and equipment distortion. Computer simulations are included to illustrate the analytical results.

An adaptive procedure for updating the coefficients of a fractionally spaced or diversity channel whitened matched filter (WMF) is presented. By exploiting the minimum phase and noise whitening properties of the WMF, a dual-multichannel estimation problem is formulated. This estimation problem is then solved using Kalman filter techniques. The WMF is thus developed directly from an estimate of the channel impulse response. Simulation results are presented to demonstrate that this technique provides a causal approximation to the WMF.>

The authors present three third- and three fourth-order cumulant based algorithms for blind deconvolution and identification of the nonminimum phase (NMP) systems. In the algorithms, based on a noncausal AR (autoregressive) model and a theorem relating to the inverse filter coefficients, the problem of blind deconvolution and identification of a NMP system is reduced to that of solving the corresponding set of linear equations. Thus, the uniqueness of the solution can normally be guaranteed. Furthermore, only the diagonal slices of cumulants are employed in the algorithms, which results in the algorithms being simpler and more accurate. A simulation example is presented for the case of unskewed continuous input, and the feasibility and efficiency of the algorithm are confirmed.>

The performance of a Kalman decision-feedback equalizer (DFE) that uses a channel estimator based on a least-mean-squares (LMS) algorithm is studied for a variety of stationary and nonstationary communications channels. This structure provides a means of model order reduction by using the residuals of the LMS to provide information on the unmodeled paths in the communication channel, which is then incorporated into the Kalman DFE structure as observation noise. The structure is compared with a conventional DFE that is trained by a Godard-Kalman algorithm with exponential windowing and adaptive Kalman structure previously reported (B. Mulgrew and C.F.N. Cowan, 1987). The results indicate that the best performance, in terms of final MSE (mean square error), is offered by the adaptive Kalman DFE structure, the final MSE being lower than that achieved by the conventional DFE by some 5-10 dB.>

In this paper a comparison is made between the convergence and tracking properties of Least Squares (LS) and Least Mean Squares (LMS) algorithms as high frequency (HF) channel estimators. Theoretical results are derived for the asymptotic error achieved by the LS algorithms under white-input conditions in the HF channel. This result is more accurate than previous analyses of LS algorithms in a nonstationary enviroment [5,8,9]. Utilising a state space definition of the channel model a minimum variance Kalman estimator is derived using the a-priori knowledge of the parameters which define the Markov process.

Using discrete time Wiener filtering theory a closed form for the optimum mean-square error (MSE) infinite impulse response (IIR) linear equaliser is derived. The minimum phase spectral factorisation, which is an integral part of the derivation of the IIR equaliser, may be circumvented through the use of a Kalman equaliser such as that originally proposed by Lawrence and Kaufman. The structure is made adaptive by using a system identification algorithm operating in parallel with a Kalman equaliser. In common with Luvison & Pirani, a least mean squares (LMS) algorithm was chosen for the system identification because the input to the channel is white. A new technique is introduced which both estimates the variance of channel noise and compensates the Kalman filter for errors in the estimate of the channel impulse response.

This paper presents a new Block Least Mean Squares(BLMS) adaptive filter structure by exploiting the use of an efficient circular convolution algorithm known as the rectangular transform(RT). The BLMS algorithm is presented as an approximation to a recursive block least squares (RBLS) algorithm. The analysis is used to develop the computationally superior overlap save adaptive filter structure even though the structure of the overlap add adaptive filter may be easily worked out in a similar manner. The computational requirements of the proposed filter are investigated and compared with an FFT based filter. The proposed filter is found to be a useful alternative to its FFT counterpart. For completeness, the convergence properties of the filter obtained from simulations are also included.