This paper considers matched-field tracking of a moving acoustic source in the ocean when acousti... more This paper considers matched-field tracking of a moving acoustic source in the ocean when acoustical properties of the environment (water column and seabed) are poorly known. The goal is not simply to estimate source locations but to determine track uncertainty distributions, thereby quantifying the information content of the tracking process. A Bayesian formulation [1, 2] is applied in which source and environmental parameters are considered unknown random variables constrained by noisy acoustic data and by prior information on parameter values (e.g., physical limits for environmental properties) and on inter-parameter relationships (limits on source velocity). Source information is extracted from the posterior probability density (PPD) by integrating over unknown environmental parameters to obtain a time-ordered series of joint marginal probability surfaces over source range and depth. Given the strong nonlinearity of the localization problem, marginal PPDs are computed numerically using efficient Markov-chain Monte Carlo (MCMC) methods, including Metropolis-Hastings sampling over environmental parameters (rotated into principal components and applying linearized proposal distributions) and heat-bath Gibbs sampling over source locations . The approach is illustrated here using acoustic data collected in the Mediterranean Sea, with tracking information content considered as a function of data quantity (number of time samples).
In this paper a procedure is developed using the Modified Hough Transform (MHT) for the detection... more In this paper a procedure is developed using the Modified Hough Transform (MHT) for the detection of signals of low signal-to-noise ratio. The signal can be considered as being composed of a sum of narrow lines. The expected number of signal and false lines to be found in an image can be determined as a function of the false alarm probability per decision. The MHT performance is compared to that of the matched filter. The method is applied to noisy images containing a wake-like signal. The MHT is able to detect the signal pattern at signal-to-noise ratios for which the signal is just visible on a good image display system. A two pass system, employing prior knowledge about the signal, gives improved detection performance.
Journal of Statistical Planning and Inference, 2011
A D-optimal minimax design criterion is proposed to construct two-level fractional factorial desi... more A D-optimal minimax design criterion is proposed to construct two-level fractional factorial designs, which can be used to estimate a linear model with main effects and some specified interactions. D-optimal minimax designs are robust against model misspecification and have small biases if the linear model contains more interaction terms. When the D-optimal minimax criterion is compared with the D-optimal design criterion, we find that the D-optimal design criterion is quite robust against model misspecification. Lower and upper bounds derived for the loss functions of optimal designs can be used to estimate the efficiencies of any design and evaluate the effectiveness of a search algorithm. Four algorithms to search for optimal designs for any run size are discussed and compared through several examples. An annealing algorithm and a sequential algorithm are particularly effective to search for optimal designs.
Some of the main features and a typical application of STARPAK (Simulation for Testing Array Resp... more Some of the main features and a typical application of STARPAK (Simulation for Testing Array Response) is presented. STARPAK is a package of Fortran subroutines designed to study the performance of an arbitrary planar array in a variety of Gaussian signal-noise environments. Array data with the appropriate statistics are simulated and then processed using the conventional, optimal and Bienvenu techniques.
A method based on the minimization of cross-entropy is presented for the recovery of signals from... more A method based on the minimization of cross-entropy is presented for the recovery of signals from noisy data either in the form of time series or images. Finite Fourier transforms are applied to the data and constraints are placed on the magnitude and phase of the Fourier coefficients based on their statistics for noise-only data. The minimization of cross-entropy is achieved through application of well-established functional minimization techniques which allow for further constraints in the spatial, temporal or frequency domain. Derivatives of the entropy function are obtained analytically and the results applied to the cases of correlated noise and of signal perturbations about a mean. Demonstrations of applications to one-dimensional data are presented.
The Journal of the Acoustical Society of America, 2015
This paper develops a matched-field approach to localization and spectral estimation of an unknow... more This paper develops a matched-field approach to localization and spectral estimation of an unknown number of ocean acoustic sources employing massively parallel implementation on a graphics processing unit (GPU) for real-time efficiency. A Bayesian formulation is developed in which the locations and complex spectra of multiple sources and noise variances are considered unknown random variables, and the Bayesian information criterion is minimized to estimate these parameters, as well as the number of sources present. Optimization is carried out using simulated annealing and includes steps that attempt to add/delete sources to/from the model. Closed-form maximum-likelihood (ML) solutions for source spectra and noise variances in terms of the source locations allow these parameters to be sampled implicitly, substantially reducing the dimensionality of the inversion. Source sampling, addition, and deletion are based on joint conditional probability distributions for source range and dep...
This paper applies a nonlinear Bayesian formulation to study uncertainty in ocean acoustic source... more This paper applies a nonlinear Bayesian formulation to study uncertainty in ocean acoustic source localization due to uncertainty in the knowledge o f ocean environmental properties (water-column sound-speed profile and seabed geoacoustic parameters). Localization uncertainty is quantified in terms o f probability ambiguity surfaces (PAS), which consist o f joint marginal probability distributions for source range and depth integrated over uncertain environmental parameters. The integration is carried out using Metropolis Gibbs' sampling for environmental parameters and two-dimensional heat-bath Gibbs' sampling for source range and depth to provide efficient sampling over complicated source search spaces with many isolated local maxima [1]. The approach is illustrated for acoustic data recorded on a hydrophone array in a shallow-water environment in the Mediterranean Sea where previous geoacoustic studies have been carried out . Localization uncertainty is considered as a function of the level o f uncertainty in the prior information for environmental properties.
This paper considers matched-field tracking of a moving acoustic source in the ocean when acousti... more This paper considers matched-field tracking of a moving acoustic source in the ocean when acoustical properties of the environment (water column and seabed) are poorly known. The goal is not simply to estimate source locations but to determine track uncertainty distributions, thereby quantifying the information content of the tracking process. A Bayesian formulation [1, 2] is applied in which source and environmental parameters are considered unknown random variables constrained by noisy acoustic data and by prior information on parameter values (e.g., physical limits for environmental properties) and on inter-parameter relationships (limits on source velocity). Source information is extracted from the posterior probability density (PPD) by integrating over unknown environmental parameters to obtain a time-ordered series of joint marginal probability surfaces over source range and depth. Given the strong nonlinearity of the localization problem, marginal PPDs are computed numerically using efficient Markov-chain Monte Carlo (MCMC) methods, including Metropolis-Hastings sampling over environmental parameters (rotated into principal components and applying linearized proposal distributions) and heat-bath Gibbs sampling over source locations . The approach is illustrated here using acoustic data collected in the Mediterranean Sea, with tracking information content considered as a function of data quantity (number of time samples).
In this paper a procedure is developed using the Modified Hough Transform (MHT) for the detection... more In this paper a procedure is developed using the Modified Hough Transform (MHT) for the detection of signals of low signal-to-noise ratio. The signal can be considered as being composed of a sum of narrow lines. The expected number of signal and false lines to be found in an image can be determined as a function of the false alarm probability per decision. The MHT performance is compared to that of the matched filter. The method is applied to noisy images containing a wake-like signal. The MHT is able to detect the signal pattern at signal-to-noise ratios for which the signal is just visible on a good image display system. A two pass system, employing prior knowledge about the signal, gives improved detection performance.
Journal of Statistical Planning and Inference, 2011
A D-optimal minimax design criterion is proposed to construct two-level fractional factorial desi... more A D-optimal minimax design criterion is proposed to construct two-level fractional factorial designs, which can be used to estimate a linear model with main effects and some specified interactions. D-optimal minimax designs are robust against model misspecification and have small biases if the linear model contains more interaction terms. When the D-optimal minimax criterion is compared with the D-optimal design criterion, we find that the D-optimal design criterion is quite robust against model misspecification. Lower and upper bounds derived for the loss functions of optimal designs can be used to estimate the efficiencies of any design and evaluate the effectiveness of a search algorithm. Four algorithms to search for optimal designs for any run size are discussed and compared through several examples. An annealing algorithm and a sequential algorithm are particularly effective to search for optimal designs.
Some of the main features and a typical application of STARPAK (Simulation for Testing Array Resp... more Some of the main features and a typical application of STARPAK (Simulation for Testing Array Response) is presented. STARPAK is a package of Fortran subroutines designed to study the performance of an arbitrary planar array in a variety of Gaussian signal-noise environments. Array data with the appropriate statistics are simulated and then processed using the conventional, optimal and Bienvenu techniques.
A method based on the minimization of cross-entropy is presented for the recovery of signals from... more A method based on the minimization of cross-entropy is presented for the recovery of signals from noisy data either in the form of time series or images. Finite Fourier transforms are applied to the data and constraints are placed on the magnitude and phase of the Fourier coefficients based on their statistics for noise-only data. The minimization of cross-entropy is achieved through application of well-established functional minimization techniques which allow for further constraints in the spatial, temporal or frequency domain. Derivatives of the entropy function are obtained analytically and the results applied to the cases of correlated noise and of signal perturbations about a mean. Demonstrations of applications to one-dimensional data are presented.
The Journal of the Acoustical Society of America, 2015
This paper develops a matched-field approach to localization and spectral estimation of an unknow... more This paper develops a matched-field approach to localization and spectral estimation of an unknown number of ocean acoustic sources employing massively parallel implementation on a graphics processing unit (GPU) for real-time efficiency. A Bayesian formulation is developed in which the locations and complex spectra of multiple sources and noise variances are considered unknown random variables, and the Bayesian information criterion is minimized to estimate these parameters, as well as the number of sources present. Optimization is carried out using simulated annealing and includes steps that attempt to add/delete sources to/from the model. Closed-form maximum-likelihood (ML) solutions for source spectra and noise variances in terms of the source locations allow these parameters to be sampled implicitly, substantially reducing the dimensionality of the inversion. Source sampling, addition, and deletion are based on joint conditional probability distributions for source range and dep...
This paper applies a nonlinear Bayesian formulation to study uncertainty in ocean acoustic source... more This paper applies a nonlinear Bayesian formulation to study uncertainty in ocean acoustic source localization due to uncertainty in the knowledge o f ocean environmental properties (water-column sound-speed profile and seabed geoacoustic parameters). Localization uncertainty is quantified in terms o f probability ambiguity surfaces (PAS), which consist o f joint marginal probability distributions for source range and depth integrated over uncertain environmental parameters. The integration is carried out using Metropolis Gibbs' sampling for environmental parameters and two-dimensional heat-bath Gibbs' sampling for source range and depth to provide efficient sampling over complicated source search spaces with many isolated local maxima [1]. The approach is illustrated for acoustic data recorded on a hydrophone array in a shallow-water environment in the Mediterranean Sea where previous geoacoustic studies have been carried out . Localization uncertainty is considered as a function of the level o f uncertainty in the prior information for environmental properties.
Uploads
Papers by Michael Wilmut