Saeed A. Aldosari

Topology for Distributed Inference on Graphs

Let N decision-makers collaborate to reach a decision. We consider iterative distributed inference with local intersensor communication, which, under simplifying assumptions, is equivalent to distributed average consensus. We show that, under appropriate conditions, the topology given by the nonbipartite Ramanujan graphs optimizes the convergence rate of this distributed algorithm.

Fusion in sensor networks with communication constraints

We address the problem of optimizing the detection performance of sensor networks under communication constraints on the common access channel. Our work helps understanding tradeoffs between sensor network parameters like number of sensors, degree of quantization at each local sensor, and SNR. Traditionally, this problem is tackled using asymptotic assumptions on the number of sensors, an approach that leads to the abstraction of important details such as the structure of the fusion center. We adopt a non-asymptotic approach and optimize both, the sensing and the fusion sides with respect to the probability of detection error. We show that the optimal fusion rule has an interesting structure similar to the majority-voting rule. In addition, we study the convergence with respect to the number of sensors of the performance of the fusion rule. We show that convergence is SNR dependent and that, in low-SNR environments, asymptotics may require a large number of sensors.

A New Low SAR Antenna Structure for Wireless Handset Applications

This paper proposes a new mobile handset antenna structure to reduce the value of the speciflc absorption rate (SAR). The antenna is based on the PIFA structure and operates at dual-bands of 0.9GHz and 1.8GHz. The chassis current is reduced using a metallic shim-layer inserted between the patch and chassis. This shim-layer is connected to the handset chassis through posts whose number and positions are determined using optimization techniques. Sidewalls are attached to increase the gain of the antenna and reduce the radiation towards human head. Simulations in the cheek mode show that the SAR reduction factor (SRF) of the proposed structure averaged over 10-g is more than 75% at 0.9GHz and 46% at 1.8GHz. The SRF values obtained using simulations and measurements are found to be better than 51% and 76% at 0.9GHz and 1.8GHz, respectively.

Detection in Sensor Networks: The Saddlepoint Approximation

This paper presents a computationally simple and accurate method to compute the error probabilities in decentralized detection in sensor networks. The cost of the direct computation of these probabilities-e.g., the probability of false alarm, the probability of a miss, or the average error probability-is combinatorial in the number of sensors and becomes infeasible even with small size networks. The method is based on the theory of large deviations, in particular, the saddlepoint approximation and applies to generic parallel fusion sensor networks, including networks with nonidentical sensors, nonidentical observations, and unreliable communication links. The paper demonstrates with parallel fusion sensor network problems the accuracy of the saddlepoint methodology: 1) computing the detection performance for a variety of small and large sensor network scenarios; and 2) designing the local detection thresholds. Elsewhere, we have used the saddlepoint approximation to study tradeoffs among parameters for networks of arbitrary size

Detection in decentralized sensor networks

Advances in integrated technologies are making networks of many inexpensive deployable autonomous sensors a reality. Individually, each sensor may not accomplish much, but working cooperatively they have for example the potential to monitor large areas, detect the presence or absence of targets, or track moving objects. These sensors operate under constraints imposed by scarce power and other limited resources like bandwidth or computing capacity. The paper considers detection in such a distributed sensor environment. We investigate the impact on detection performance, as measured by the probability of error, of such parameters as number of sensors, number of quantization levels at each sensor, or signal to noise ratio, under a rate constraint on the common access communications channel. We optimize the local detectors when the number of sensors is large. We show that the performance loss due to quantization decays exponentially fast as the number of bits per sensor increases and that the choice between hard versus soft local detectors depends not only on the noise distribution and the quantization rate, but also on the SNR under which the sensors operate.

Saddlepoint approximation for sensor network optimization

The task of detection optimization in sensor networks is hindered by the large computational cost of evaluating the performance criteria, e.g. the probability of making wrong decisions. We present an approach that avoids these obstacles by considering a rather accurate approximation to computing the detection performance. We propose the saddlepoint approximation and provide results that demonstrate its high accuracy and low complexity. The results are used to show that, for a range of problems, the optimal fusion rule is equivalent to a simple majority rule.

A new MSE approach for combined linear-Viterbi equalizers

Combined linear-Viterbi equalization (CLVE) is a technique that employs a linear pre-filter in conjunction with the Viterbi algorithm (VA) to mitigate the effects of intersymbol interference. The aim of the linear pre-filter is to shape the original channel impulse response to some shorter desired impulse response (DIR) in order to reduce the complexity of the VA. In this paper, we present a new MSE based approach for optimizing CLVEs. This approach takes advantage of the modifications to the VA which are suitable for channels having coarsely located coefficients. Specifically, the new approach has the flexibility in choosing the positions and optimizing the values of nonzero coefficients of DIR. As a result, it includes the conventional MSE-based approaches as a special case. Simulation results have been presented to illustrate the performance of proposed method.

MEG data classification for healthy and epileptic subjects using linear discriminant analysis

Electroencephalogram (EEG) is the most commonly used clinical tool for the early diagnosis of epilepsy. However, with the recent advances in the magnetoencephalography (MEG) technology, a new source of information for the analysis of brain signals has been established. Epileptologists often spend considerable amount of time to review MEG recordings to determine whether or not a particular subject can be classified as an epileptic patient. This paper proposes a new algorithm for automatic classification of MEG data into two classes: data that belongs to healthy subjects and data that belongs to epileptic subjects. The classifier makes use of linear discriminant analysis (LDA) and considers features extracted from the signals of eight regions in the brain. The effectiveness of proposed classifier has been tested using real MEG data obtained from 15 healthy subjects and 18 epilepsy patients. The results obtained show good promise, which make the proposed classifier a valuable tool for analyzing brain signals in the initial assessment phases of subjects under epileptic symptoms.

Combined linear-decision feedback sequence estimation: an improved system design

Decision feedback sequence estimation (DFSE) is a reduced state alternative to maximum likelihood sequence estimation (MLSE). In this paper, we examine the performance of a system composed of a linear pre-filter in conjunction with a DFSE. In addition, we present a new optimization approach for the linear pre-filter. This approach has an advantage over MSE-based approaches in that it takes into account the effects of noise correlation and error distance variations. The performance of the proposed optimization approach is evaluated and compared with several existing optimization techniques.

Online adaptive seizure prediction algorithm for scalp EEG

Epilepsy is a brain disorder, which affects around 1% of world population. The life of epilepsy patients can be improved by predicting seizures before its occurrence. It has been observed that EEG signals during the pre-seizure state are less chaotic compared to their behavior at normal state. Therefore, chaoticity measure can be used to develop seizure predictor. In this paper, we propose seizure prediction algorithm based on Largest Lyapunov Exponent (LLE) to measure the chaoticity of scalp EEG signals. The proposed algorithm makes use of LLE to define two baselines; one for the normal state and the other for the pre-state. The distance between the two baselines and the LLEs of an Electroencephalography (EEG) signal of unknown state is computed for signal classification. The two baselines are updated through a simple mechanism. The performance of proposed algorithm has been evaluated using MIT database.