Ashok Sundaresan

Distributed detection of a nuclear radioactive source using fusion of correlated decisions

A distributed detection method is developed for the detection of a nuclear radioactive source using a small number of radiation counters. Local one bit decisions are made at each sensor over a period of time and a fusion center makes the global decision. A novel test for the fusion of correlated decisions is derived using the theory of copulas and optimal sensor thresholds are obtained using the Normal copula function. The performance of the derived fusion rule is compared with that of the Chair-Varshney rule. An increase in detection performance is observed. A method to estimate the correlation between the sensor observations using only the vector of sensor decisions is also proposed.

Copula-Based Fusion of Correlated Decisions

Detection of random signals under a distributed setting is considered. Due to the random nature of the spatial phenomenon being observed, the sensor decisions collected at the fusion center are correlated. Assuming that local detectors are single threshold binary quantizers, a novel approach for the fusion of correlated decisions is proposed using the theory of copulas. The proposed approach assumes only the knowledge of the marginal distribution of sensor observations but no prior knowledge of their joint distribution. Using a Neyman-Pearson (NP) framework for detection at the fusion center, the optimal fusion rule is derived. An example involving the detection of nuclear radiation is presented to illustrate the proposed approach, and results demonstrating the efficiency of the copula-based fusion rule are shown.

Location Estimation of a Random Signal Source Based on Correlated Sensor Observations

The problem of location estimation of a source of random signals using a network of sensors is considered. A novel maximum-likelihood estimation (MLE) based approach using copula functions is proposed. The measurements received at the sensors are often spatially correlated and characterized by a multivariate distribution. Using the theory of copulas, the joint parametric density of sensor observations (joint likelihood) is approximated assuming only the knowledge of the marginal likelihood functions of the sensor observations. The problem of selecting the best copula function to model the joint likelihood is approached as one of model selection and a model fusion strategy is used to reduce the effect of selection bias. An example involving source localization of a Poisson source is presented to illustrate the proposed approach and demonstrate its performance.

Robustness of Change Detection Algorithms in the Presence of Registration Errors

Accurate registration of multi-temporal remote sensing images is critical to any change detection study. The presence of registration errors in the images may affect the accuracy of change detection. In this paper, we evaluate the performance of two change detection algorithms in the presence of artificially introduced registration errors in the dataset. The algorithms considered are image differencing and an algorithm based on a Markov random field (MRF) model. Registration errors have been introduced in four different ways: only in x direction, only in y direction, in both x and y directions without any rotational misregistration, and finally in both x and y directions together with rotational misregistration. Three temporal datasets, a simulated dataset and two synthetic datasets created from remote sensing images acquired by the Landsat TM sensor, have been used in our study. The results indicate that the change detection algorithm based on the MRF model is more robust to the presence of registration errors than the image differencing method.

A copula-based semi-parametric approach for footstep detection using seismic sensor networks

In this paper, we consider the problem of detecting the presence of footsteps using signal measurements from a network of seismic sensors. Since the sensors are closely spaced, they result in correlated measurements. A novel method for detection that exploits the spatial dependence of sensor measurements using copula functions is proposed. An approach for selecting the copula function that is most suited for modeling the spatial dependence of sensor observations is also provided. The performance of the proposed approach is illustrated using real footstep signals collected using an experimental test-bed consisting of seismic sensors.

Detection of Dependent Heavy-Tailed Signals

This paper examines the problem of detection of dependent α-stable signals. Measurements of several phenomena exhibit non-Gaussian, heavy-tailed behavior in their probability density functions (p.d.f.); we use the class of α-stable distributions to characterize these signals. When two sensors make simultaneous measurements of such phenomena, these heavy-tailed realizations are dependent across sensors. The intersensor dependence is modeled using copulas. We consider a two-sided test in the Neyman-Pearson framework and present an asymptotic analysis of the generalized likelihood test (GLRT). Both, nested and non-nested models are considered in the analysis. The performance of the proposed scheme is evaluated numerically on simulated data, as well as indoor seismic data. With appropriately selected models, our results demonstrate that a high probability of detection can be achieved for false alarm probabilities of the order of 10-4.

Distributed detection of a nuclear radioactive source based on a hierarchical source model

Detection of a nuclear radioactive source is considered using a parallel sensor network architecture and a fusion center. A Poisson-Gamma hierarchical model is used to represent the distribution of the count data received by the sensors. Local sensors are assumed to be single threshold binary quantizers that send a vector of sensor decisions over time to the fusion center for global decision-making. Using the developed count model, a generalized likelihood ratio test (GLRT) using a restricted range MLE (RMLE) is proposed to declare the global decision. The performance improvement resulting from using the restricted range MLE over the unrestricted MLE while implementing the GLRT is depicted using simulated as well as real data collected from a test-bed using radiation sensors. Using bootstrap, 95% confidence bounds on the ROC curves, evaluated using real data, are obtained.

Data processing in multivariable RFID vapor sensors

Sensors for selective monitoring of gases and volatiles are needed for numerous applications including medical diagnostics, food safety, environmental, industrial, homeland protection, and many others. For these and other applications, we have developed passive radio frequency identification (RFID) sensors for vapor sensing where we apply a sensing film onto the resonant antenna of the RFID sensor, simultaneously measure several parameters of antenna impedance, and process these parameters using multivariate analysis tools. In this work, we critically analyze techniques of processing the impedance response of individual sensors coated with different sensing materials and the ability of these techniques to increase selectivity of developed sensors upon exposure to model vapors. Four types of investigated data processing techniques are based on unsupervised and supervised pattern recognition algorithms. Two evaluation criteria for these techniques involved their ability (1) to correctly identify types of vapors and (2) to provide the smallest error of prediction of concentrations of vapors.

On localizing the source of random signals using sensor networks

The problem of source localization using a network of sensors is considered. A maximum likelihood estimation (MLE) based approach is adopted. The measurements received at the sensors due to the random phenomenon are spatially correlated and are characterized by a multivariate distribution. Using the theory of copulas, the joint parametric density of sensor observations is obtained assuming only the knowledge of their marginal densities. An example showing the efficiency of the proposed approach is presented.