Publications

Export 155 results:
Sort by: Author Title Type [ Year  (Desc)]
2019
Nannuru, S, Gemba KL, Gerstoft P, Hodgkiss WS, Mecklenbrauker CF.  2019.  Sparse Bayesian learning with multiple dictionaries. Signal Processing. 159:159-170.   10.1016/j.sigpro.2019.02.003   AbstractWebsite

Sparse Bayesian learning (SBL) has emerged as a fast and competitive method to perform sparse processing. The SBL algorithm, which is developed using a Bayesian framework, iteratively solves a non-convex optimization problem using fixed point updates. It provides comparable performance and is significantly faster than convex optimization techniques used in sparse processing. We propose a multi-dictionary SBL algorithm that simultaneously can process observations generated by different underlying dictionaries sharing the same sparsity profile. Two algorithms are proposed and corresponding fixed point update equations are derived. Noise variances are estimated using stochastic maximum likelihood. The multi dictionary SBL has many practical applications. We demonstrate this using direction-of-arrival (DOA) estimation. The first example uses the proposed multi-dictionary SBL to process multi-frequency observations. We show how spatial aliasing can be avoided while processing multi-frequency observations using SBL. SWellEx-96 experimental data demonstrates qualitatively these advantages. In the second example we show how data corrupted with heteroscedastic noise can be processed using multi-dictionary SBL with data pre-whitening. (C) 2019 Elsevier B.V. All rights reserved.

White-Gaynor, AL, Nyblade AA, Aster RC, Wiens DA, Bromirski PD, Gerstoft P, Stephen RA, Hansen SE, Wilson T, Dalziel IW, Huerta AD, Winberry JP, Anandakrishnan S.  2019.  Heterogeneous upper mantle structure beneath the Ross Sea Embayment and Marie Byrd Land, West Antarctica, revealed by P-wave tomography. Earth and Planetary Science Letters. 513:40-50.   10.1016/j.epsl.2019.02.013   AbstractWebsite

We present an upper mantle P-wave velocity model for the Ross Sea Embayment (RSE) region of West Antarctica, constructed by inverting relative P-wave travel-times from 1881 teleseismic earthquakes recorded by two temporary broadband seismograph deployments on the Ross Ice Shelf, as well as by regional ice- and rock-sited seismic stations surrounding the RSE. Faster upper mantle P-wave velocities (similar to +1%) characterize the eastern part of the RSE, indicating that the lithosphere in this part of the RSE may not have been reheated by mid-to-late Cenozoic rifting that affected other parts of the Late Cretaceous West Antarctic Rift System. Slower upper mantle velocities (similar to -1%) characterize the western part of the RSE over a similar to 500 km-wide region, extending from the central RSE to the Transantarctic Mountains (TAM). Within this region, the model shows two areas of even slower velocities (similar to -1.5%) centered beneath Mt. Erebus and Mt. Melbourne along the TAM front. We attribute the broader region of slow velocities mainly to reheating of the lithospheric mantle by Paleogene rifting, while the slower velocities beneath the areas of recent volcanism may reflect a Neogene-present phase of rifting and/or plume activity associated with the formation of the Terror Rift. Beneath the Ford Ranges and King Edward VII Peninsula in western Marie Byrd Land, the P-wave model shows lateral variability in upper mantle velocities of +/- 0.5% over distances of a few hundred km. The heterogeneity in upper mantle velocities imaged beneath the RSE and western Marie Byrd Land, assuming no significant variation in mantle composition, indicates variations in upper mantle temperatures of at least 100 degrees C. These temperature variations could lead to differences in surface heat flow of similar to +/- 10 mW/m(2) and mantle viscosity of 10(2) Pa s regionally across the study area, possibly influencing the stability of the West Antarctic Ice Sheet by affecting basal ice conditions and glacial isostatic adjustment. (C) 2019 Elsevier B.V. All rights reserved.

Park, Y, Gerstoft P, Seong W.  2019.  Grid-free compressive mode extraction. Journal of the Acoustical Society of America. 145:1427-1442.   10.1121/1.5094345   AbstractWebsite

A grid-free compressive sensing (CS) based method for extracting the normal modes of acoustic propagation in the ocean waveguide from vertical line array (VLA) data is presented. Extracting the normal modes involves the estimation of mode horizontal wavenumbers and the corresponding mode shapes. Sparse representation of the waveguide propagation using modes at discrete horizontal wavenumbers enables CS to be applied. Grid-free CS, based on group total-variation norm minimization, is adopted to mitigate the issues of the wavenumber search grid discretization in the conventional CS. In addition, the suggested method can process multiple sensor data jointly, which improves performance in estimation over single sensor data processing. The method here uses data on a VLA from a source at several ranges, and processes the multiple sensor data at different depths jointly. The grid-free CS extracts the mode wavenumbers and shapes even with no a priori environmental knowledge, a partial water column spanning array data, and without the mode orthogonality condition. The approach is illustrated by numerical simulations and experimental SWellEx-96 (shallow water evaluation cell experiment 1996) data. (C) 2019 Acoustical Society of America.

Gemba, KL, Nannuru S, Gerstoft P.  2019.  Robust ocean acoustic localization with sparse Bayesian learning. Ieee Journal of Selected Topics in Signal Processing. 13:49-60.   10.1109/jstsp.2019.2900912   AbstractWebsite

Matched field processing (MFP) compares the measures to the modeled pressure fields received at an array of sensors to localize a source in an ocean waveguide. Typically, there are only a few sources when compared to the number of candidate source locations or range-depth cells. We use sparse Bayesian learning (SBL) to learn a common sparsity profile corresponding to the location of present sources. SBL performance is compared to traditional processing in simulations and using experimental ocean acoustic data. Specifically, we localize a quiet source in the presence of a surface interferer in a shallow water environment. This multi-frequency scenario requires adaptive processing and includes modest environmental and sensor position mismatch in the MFP model. The noise process changes likely with time and is modeled as a non-stationary Gaussian process, meaning that the noise variance changes across snapshots. The adaptive SBL algorithm models the complex source amplitudes as random quantities, providing robustness to amplitude and phase errors in the model. This is demonstrated with experimental data, where SBL exhibits improved source localization performance when compared to the white noise gain constraint (-3 dB) and Bartlett processors.

Kong, QK, Trugman DT, Ross ZE, Bianco MJ, Meade BJ, Gerstoft P.  2019.  Machine learning in seismology: Turning data into insights. Seismological Research Letters. 90:3-14.   10.1785/0220180259   AbstractWebsite

This article provides an overview of current applications of machine learning (ML) in seismology. ML techniques are becoming increasingly widespread in seismology, with applications ranging from identifying unseen signals and patterns to extracting features that might improve our physical understanding. The survey of the applications in seismology presented here serves as a catalyst for further use of ML. Five research areas in seismology are surveyed in which ML classification, regression, clustering algorithms show promise: earthquake detection and phase picking, earthquake early warning (EEW), ground-motion prediction, seismic tomography, and earthquake geodesy. We conclude by discussing the need for a hybrid approach combining data-driven ML with traditional physical modeling.

2018
Bianco, MJ, Gerstoft P.  2018.  Travel time tomography with adaptive dictionaries. Ieee Transactions on Computational Imaging. 4:499-511.   10.1109/tci.2018.2862644   AbstractWebsite

We develop a two-dimensional travel time tomography method, which regularizes the inversion by modeling groups of slowness pixels from discrete slowness maps, called patches, as sparse linear combinations of atoms from a dictionary. We propose to use dictionary learning during the inversion to adapt dictionaries to specific slowness maps. This patch regularization, called the local model, is integrated into the overall slowness map, called the global model. The local model considers small-scale variations using a sparsity constraint, and the global model considers larger-scale features constrained using l(2) regularization. This strategy in a locally sparse travel time tomography (LST) approach enables simultaneous modeling of smooth and discontinuous slowness features. This is in contrast to conventional tomography methods, which constrain models to be exclusively smooth or discontinuous. We develop a maximum a posteriori formulation for LST and exploit the sparsity of slowness patches using dictionary learning. The LST approach compares favorably with smoothness and total variation regularization methods on densely, but irregularly sampled synthetic slowness maps.

Nannuru, S, Koochakzadeh A, Gemba KL, Pal P, Gerstoft P.  2018.  Sparse Bayesian learning for beamforming using sparse linear arrays. The Journal of the Acoustical Society of America. 144:2719-2729.   10.1121/1.5066457   Abstract

Sparse linear arrays such as co-prime and nested arrays can resolve more sources than the number of sensors. In contrast, uniform linear arrays (ULA) cannot resolve more sources than the number of sensors. This paper demonstrates this using Sparse Bayesian learning (SBL) and co-array MUSIC for single frequency beamforming. For approximately the same number of sensors, co-prime and nested arrays are shown to outperform ULA in root mean squared error. This paper shows that multi-frequency SBL can significantly reduce spatial aliasing. The effects of different sparse sub-arrays on SBL performance are compared qualitatively using the Noise Correlation 2009 experimental data set.

Chaput, J, Aster RC, McGrath D, Baker M, Anthony RE, Gerstoft P, Bromirski P, Nyblade A, Stephen RA, Wiens DA, Das SB, Stevens LA.  2018.  Near-surface environmentally forced changes in the Ross Ice Shelf observed with ambient seismic noise. Geophysical Research Letters. 45:11187-11196.   10.1029/2018gl079665   AbstractWebsite

Continuous seismic observations across the Ross Ice Shelf reveal ubiquitous ambient resonances at frequencies >5 Hz. These firn-trapped surface wave signals arise through wind and snow bedform interactions coupled with very low velocity structures. Progressive and long-term spectral changes are associated with surface snow redistribution by wind and with a January 2016 regional melt event. Modeling demonstrates high spectral sensitivity to near-surface (top several meters) elastic parameters. We propose that spectral peak changes arise from surface snow redistribution in wind events and to velocity drops reflecting snow lattice weakening near 0 degrees C for the melt event. Percolation-related refrozen layers and layer thinning may also contribute to long-term spectral changes after the melt event. Single-station observations are inverted for elastic structure for multiple stations across the ice shelf. High-frequency ambient noise seismology presents opportunities for continuous assessment of near-surface ice shelf or other firn environments. Plain Language Summary Ice shelves are the floating buttresses of large glaciers that extend over the oceans and play a key role in restraining inland glaciers as they flow to the sea. Deploying sensitive seismographs across Earth's largest ice shelf (the Ross Ice Shelf) for 2 years, we discovered that the shelf nearly continuously sings at frequencies of five or more cycles per second, excited by local and regional winds blowing across its snow dune-like topography. We find that the frequencies and other features of this singing change, both as storms alter the snow dunes and during a (January 2016) warming event that resulted in melting in the ice shelf's near surface. These observations demonstrate that seismological monitoring can be used to continually monitor the near-surface conditions of an ice shelf and other icy bodies to depths of several meters.

Chen, Z, Bromirski PD, Gerstoft P, Stephen RA, Wiens DA, Aster RC, Nyblade AA.  2018.  Ocean-excited plate waves in the Ross and Pine Island Glacier ice shelves. Journal of Glaciology. 64:730-744.   10.1017/jog.2018.66   AbstractWebsite

Ice shelves play an important role in buttressing land ice from reaching the sea, thus restraining the rate of grounded ice loss. Long-period gravity-wave impacts excite vibrations in ice shelves that can expand pre-existing fractures and trigger iceberg calving. To investigate the spatial amplitude variability and propagation characteristics of these vibrations, a 34-station broadband seismic array was deployed on the Ross Ice Shelf (RIS) from November 2014 to November 2016. Two types of ice-shelf plate waves were identified with beamforming: flexural-gravity waves and extensional Lamb waves. Below 20 mHz, flexural-gravity waves dominate coherent signals across the array and propagate landward from the ice front at close to shallow-water gravity-wave speeds (similar to 70 m s(-1)). In the 20-100 mHz band, extensional Lamb waves dominate and propagate at phase speeds similar to 3 km s(-1). Flexural-gravity and extensional Lamb waves were also observed by a 5-station broadband seismic array deployed on the Pine Island Glacier (PIG) ice shelf from January 2012 to December 2013, with flexural wave energy, also detected at the PIG in the 20-100 mHz band. Considering the ubiquitous presence of storm activity in the Southern Ocean and the similar observations at both the RIS and the PIG ice shelves, it is likely that most, if not all, West Antarctic ice shelves are subjected to similar gravity-wave excitation.

Shen, WS, Wiens DA, Anandakrishnan S, Aster RC, Gerstoft P, Bromirski PD, Hansen SE, Dalziel IWD, Heeszel DS, Huerta AD, Nyblade AA, Stephen R, Wilson TJ, Winberry JP.  2018.  The crust and upper mantle structure of central and west Antarctica from bayesian inversion of rayleigh wave and receiver functions. Journal of Geophysical Research-Solid Earth. 123:7824-7849.   10.1029/2017jb015346   AbstractWebsite

We construct a new seismic model for central and West Antarctica by jointly inverting Rayleigh wave phase and group velocities along with P wave receiver functions. Ambient noise tomography exploiting data from more than 200 seismic stations deployed over the past 18years is used to construct Rayleigh wave phase and group velocity dispersion maps. Comparison between the ambient noise phase velocity maps with those constructed using teleseismic earthquakes confirms the accuracy of both results. These maps, together with P receiver function waveforms, are used to construct a new 3-D shear velocity (Vs) model for the crust and uppermost mantle using a Bayesian Monte Carlo algorithm. The new 3-D seismic model shows the dichotomy of the tectonically active West Antarctica (WANT) and the stable and ancient East Antarctica (EANT). In WANT, the model exhibits a slow uppermost mantle along the Transantarctic Mountains (TAMs) front, interpreted as the thermal effect from Cenozoic rifting. Beneath the southern TAMs, the slow uppermost mantle extends horizontally beneath the traditionally recognized EANT, hypothesized to be associated with lithospheric delamination. Thin crust and lithosphere observed along the Amundsen Sea coast and extending into the interior suggest involvement of these areas in Cenozoic rifting. EANT, with its relatively thick and cold crust and lithosphere marked by high Vs, displays a slower Vs anomaly beneath the Gamburtsev Subglacial Mountains in the uppermost mantle, which we hypothesize may be the signature of a compositionally anomalous body, perhaps remnant from a continental collision.

Wang, WT, Gerstoft P, Wang BS.  2018.  Interference of teleseismic body waves in noise cross-correlation functions in Southwest China. Seismological Research Letters. 89:1817-1825.   10.1785/0220180139   AbstractWebsite

Vertical-vertical component noise cross-correlation functions (NCFs) from southwest China show strong signals at 4-8 s periods with apparent velocities exceeding 18 km/s. These signals exhibit clear azimuthal dependency and undergo seasonal variations. Using NCF-based beamforming, we isolate these signals from surface waves to better address their slowness distributions and source locations. They are identified as the interference of teleseismic P-wave microseisms. Backprojection results indicate that the sources are from deep oceans, consistent with previous studies. The inferred source locations are verified by the agreement between observed signals and their predicted arrival times. Our study suggests that knowledge of microseism sources helps identify various signals in NCFs and that analyzing these signals improves our understanding of the noise sources.

Wang, WT, Gerstoft P, Wang BS.  2018.  Seasonality of P wave microseisms from NCF-based beamforming using ChinArray. Geophysical Journal International. 213:1832-1848.   10.1093/gji/ggy081   AbstractWebsite

Teleseismic P wave microseisms produce interference signals with high apparent velocity in noise cross-correlation functions (NCFs). Sources of P wave microseisms can be located with NCFs from seismic arrays. Using the vertical-vertical component NCFs from a large-aperture array in southwestern China (ChinArray), we studied the P wave source locations and their seasonality of microseisms at two period bands (8-12 and 4-8 s) with an NCF-based beamforming method. The sources of P, PP and PKPbc waves are located. The ambiguity between P and PP source locations is analysed using averaged significant ocean wave height and sea surface pressure as constraints. The results indicate that the persistent P wave sources are mainly located in the deep oceans such as the North Atlantic, North Pacific and Southern Ocean, in agreement with previous studies. The Gulf of Alaska is found to generate P waves favouring the 8-12 s period band. The seasonality of P wave sources is consistent with the hemispheric storm pattern, which is stronger in local winter. Using the identified sources, arrival times of the interference signals are predicted and agree well with observations. The interference signals exhibit seasonal variation, indicating that body wave microseisms in southwestern China are from multiple seasonal sources.

Siderius, M, Li J, Gerstoft P.  2018.  Head waves in ocean acoustic ambient noise: Measurements and modeling. Journal of the Acoustical Society of America. 143:1182-1193.   10.1121/1.5024332   AbstractWebsite

Seismic interferometry recovers the Green's function between two receivers by cross-correlating the field measured from sources that surround the receivers. In the seismic literature, it has been widely reported that this processing can produce artifacts in the Green's function estimate called "spurious multiples" or the "virtual refracted wave." The spurious multiples are attributed to the head wave and its multiples and travels in the seabed. The head wave phenomenon is shown to be observable from both controlled active sources and from ocean ambient noise and for both vertical and horizontal arrays. The processing used is a generalization of the passive fathometer to produce cross-beam correlations. This passive fathometer is equivalent to the seismic interferometry techniques for delay and sum beamforming but not for adaptive beamforming. Modeling and experimental data show the head wave is observed in ocean noise and can be used to estimate the seabed sound speed. (C) 2018 Acoustical Society of America.

2017
Wagner, M, Nannuru S, Gerstoft P.  2017.  Compressive MIMO beamforming of data collected in a refractive environment. Radio Science. 52:1458-1471.   10.1002/2017rs006473   AbstractWebsite

The phenomenon of ducting is caused by abnormal atmospheric refractivity patterns and is known to allow electromagnetic waves to propagate over the horizon with unusually low propagation loss. It is unknown what effect ducting has on multiple input multiple output (MIMO) channels, particularly its effect on multipath propagation in MIMO channels. A high-accuracy angle-of-arrival and angle-of-departure estimation technique for MIMO communications, which we will refer to as compressive MIMO beamforming, was tested on simulated data then applied to experimental data taken from an over the horizon MIMO test bed located in a known ducting hot spot in Southern California. The multipath channel was estimated from the receiver data recorded over a period of 18 days, and an analysis was performed on the recorded data. The goal is to observe the evolution of the MIMO multipath channel as atmospheric ducts form and dissipate to gain some understanding of the behavior of channels in a refractive environment. This work is motivated by the idea that some multipath characteristics of MIMO channels within atmospheric ducts could yield important information about the duct. Plain Language Summary Long-range ship to ship wireless communication is difficult because the horizon can obstruct the line of sight path between ships, causing radio signal strength to decrease rapidly with range. Sometimes, however, an event known as ducting can occur which allows radio waves to curve over the horizon. Multiple input multiple output (MIMO) radio setups can exploit knowledge of the paths taken by the signal from transmitter to receiver to increase communication strength. In this paper we take measurements from a MIMO radio setup located in a region known for ducting and observe the evolution of the signal paths, looking for patterns that may be used to predict properties of atmospheric ducts.

Niu, HQ, Ozanich E, Gerstoft P.  2017.  Ship localization in Santa Barbara Channel using machine learning classifiers. Journal of the Acoustical Society of America. 142:EL455-EL460.   10.1121/1.5010064   AbstractWebsite

Machine learning classifiers are shown to outperform conventional matched field processing for a deep water (600m depth) ocean acoustic-based ship range estimation problem in the Santa Barbara Channel Experiment when limited environmental information is known. Recordings of three different ships of opportunity on a vertical array were used as training and test data for the feed-forward neural network and support vector machine classifiers, demonstrating the feasibility of machine learning methods to locate unseen sources. The classifiers perform well up to 10km range whereas the conventional matched field processing fails at about 4 km range without accurate environmental information. (C) 2017 Acoustical Society of America

Ozanich, E, Gerstoft P, Worcester PF, Dzieciuch MA, Thode A.  2017.  Eastern Arctic ambient noise on a drifting vertical array. Journal of the Acoustical Society of America. 142:1997-2006.   10.1121/1.5006053   AbstractWebsite

Ambient noise in the eastern Arctic was studied from April to September 2013 using a 22 element vertical hydrophone array as it drifted from near the North Pole (89 degrees 23'N, 62 degrees 35'W) to north of Fram Strait (83 degrees 45'N, 4 degrees 28'W). The hydrophones recorded for 108 min/day on six days per week with a sampling rate of 1953.125 Hz. After removal of data corrupted by non-acoustic transients, 19 days throughout the transit period were analyzed. Noise contributors identified include broadband and tonal ice noises, bowhead whale calling, seismic airgun surveys, and earthquake T phases. The bowhead whale or whales detected are believed to belong to the endangered Spitsbergen population, and were recorded when the array was as far north as 86 degrees 24'N. Median power spectral estimates and empirical probability density functions along the array transit show a change in the ambient noise levels corresponding to seismic survey airgun occurrence and received level at low frequencies and transient ice noises at high frequencies. Median power for the same periods across the array shows that this change is consistent in depth. The median ambient noise for May 2013 was among the lowest of the sparse reported observations in the eastern Arctic but comparable to the more numerous observations of western Arctic noise levels. (C) 2017 Acoustical Society of America.

Niu, HQ, Reeves E, Gerstoft P.  2017.  Source localization in an ocean waveguide using supervised machine learning. Journal of the Acoustical Society of America. 142:1176-1188.   10.1121/1.5000165   AbstractWebsite

Source localization in ocean acoustics is posed as a machine learning problem in which data-driven methods learn source ranges directly from observed acoustic data. The pressure received by a vertical linear array is preprocessed by constructing a normalized sample covariance matrix and used as the input for three machine learning methods: feed-forward neural networks (FNN), support vector machines (SVM), and random forests (RF). The range estimation problem is solved both as a classification problem and as a regression problem by these three machine learning algorithms. The results of range estimation for the Noise09 experiment are compared for FNN, SVM, RF, and conventional matched-field processing and demonstrate the potential of machine learning for underwater source localization. (C) 2017 Acoustical Society of America.

Bromirski, PD, Chen Z, Stephen RA, Gerstoft P, Arcas D, Diez A, Aster RC, Wiens DA, Nyblade A.  2017.  Tsunami and infragravity waves impacting Antarctic ice shelves. Journal of Geophysical Research-Oceans. 122:5786-5801.   10.1002/2017jc012913   AbstractWebsite

The responses of the Ross Ice Shelf (RIS) to the 16 September 2015 8.3 (M-w) Chilean earthquake tsunami (>75 s period) and to oceanic infragravity (IG) waves (50-300 s period) were recorded by a broadband seismic array deployed on the RIS from November 2014 to November 2016. Here we show that tsunami and IG-generated signals within the RIS propagate at gravity wave speeds (similar to 70 m/s) as water-ice coupled flexural-gravity waves. IG band signals show measureable attenuation away from the shelf front. The response of the RIS to Chilean tsunami arrivals is compared with modeled tsunami forcing to assess ice shelf flexural-gravity wave excitation by very long period (VLP; >300 s) gravity waves. Displacements across the RIS are affected by gravity wave incident direction, bathymetry under and north of the shelf, and water layer and ice shelf thicknesses. Horizontal displacements are typically about 10 times larger than vertical displacements, producing dynamical extensional motions that may facilitate expansion of existing fractures. VLP excitation is continuously observed throughout the year, with horizontal displacements highest during the austral winter with amplitudes exceeding 20 cm. Because VLP flexural-gravity waves exhibit no discernable attenuation, this energy must propagate to the grounding zone. Both IG and VLP band flexural-gravity waves excite mechanical perturbations of the RIS that likely promote tabular iceberg calving, consequently affecting ice shelf evolution. Understanding these ocean-excited mechanical interactions is important to determine their effect on ice shelf stability to reduce uncertainty in the magnitude and rate of global sea level rise. Plain Language Summary A major source of the uncertainty in the magnitude and rate of global sea level rise is the contribution from Antarctica. Ice shelves buttress land ice, restraining land ice from reaching the sea. We present the analysis of seismic data collected with a broadband seismic array deployed on the Ross Ice Shelf, Antarctica. The characteristics of ocean gravity-wave-induced vibrations, that may expand existing fractures in the ice shelf and/or trigger iceberg calving or ice shelf collapse events, are described. The mechanical dynamic strains induced can potentially affect ice shelf integrity, and ultimately reduce or remove buttressing restraints, accelerating sea level rise.

Gemba, KL, Nannuru S, Gerstoft P, Hodgkiss WS.  2017.  Multi-frequency sparse Bayesian learning for robust matched field processing. Journal of the Acoustical Society of America. 141:3411-3420.   10.1121/1.4983467   AbstractWebsite

The multi-snapshot, multi-frequency sparse Bayesian learning (SBL) processor is derived and its performance compared to the Bartlett, minimum variance distortionless response, and white noise constraint processors for the matched field processing application. The two-source model and data scenario of interest includes realistic mismatch implemented in the form of array tilt and data snapshots not exactly corresponding to the range-depth grid of the replica vectors. Results demonstrate that SBL behaves similar to an adaptive processor when localizing a weaker source in the presence of a stronger source, is robust to mismatch, and exhibits improved localization performance when compared to the other processors. Unlike the basis or matching pursuit methods, SBL automatically determines sparsity and its solution can be interpreted as an ambiguity surface. Because of its computational efficiency and performance, SBL is practical for applications requiring adaptive and robust processing. (C) 2017 Acoustical Society of America.

Das, A, Hodgkiss WS, Gerstoft P.  2017.  Coherent multipath direction-of-arrival resolution using compressed sensing. Ieee Journal of Oceanic Engineering. 42:494-505.   10.1109/joe.2016.2576198   AbstractWebsite

For a sound field observed on a sensor array, performance of conventional high-resolution adaptive beamformers is affected dramatically in the presence of coherent multipath signals, but the directions-of-arrival (DOAs) and power levels of these arrivals can be resolved with compressed sensing (CS). When the number of multipath signals is sufficiently small, a CS approach can be used by formulating the problem as a sparse signal recovery problem. CS overcomes the difficulty of resolving coherent arrivals at an array by directly processing the sensor outputs without first estimating a sensor covariance matrix. CS is compared to the adaptive minimum-variance-distortionless-response (MVDR) spatial processor with spatial smoothing. Though spatial smoothing produces improved results by preprocessing the sensor array covariance matrix to decorrelate the coherent multipath components, it reduces the effective aperture of the array and hence reduces the resolution. An empirical study with a uniform linear array (ULA) demonstrates that CS outperforms MVDR beamformer with spatial smoothing in terms of spatial resolution and bias and variance of DOA and power estimates. Analysis of the shallow-water HF97 ocean acoustic experimental data shows that CS is able to recover the DOAs and power levels of the multipath signals with superior resolution compared to MVDR with spatial smoothing.

Bianco, M, Gerstoft P.  2017.  Dictionary learning of sound speed profiles. Journal of the Acoustical Society of America. 141:1749-1758.   10.1121/1.4977926   AbstractWebsite

To provide constraints on the inversion of ocean sound speed profiles (SSPs), SSPs are often modeled using empirical orthogonal functions (EOFs). However, this regularization, which uses the leading order EOFs with a minimum-energy constraint on the coefficients, often yields low resolution SSP estimates. In this paper, it is shown that dictionary learning, a form of unsupervised machine learning, can improve SSP resolution by generating a dictionary of shape functions for sparse processing (e.g., compressive sensing) that optimally compress SSPs; both minimizing the reconstruction error and the number of coefficients. These learned dictionaries (LDs) are not constrained to be orthogonal and thus, fit the given signals such that each signal example is approximated using few LD entries. Here, LDs describing SSP observations from the HF-97 experiment and the South China Sea are generated using the K-SVD algorithm. These LDs better explain SSP variability and require fewer coefficients than EOFs, describing much of the variability with one coefficient. Thus, LDs improve the resolution of SSP estimates with negligible computational burden. (C) 2017 Acoustical Society of America.

Tollefsen, D, Gerstoft P, Hodgkiss WS.  2017.  Multiple-array passive acoustic source localization in shallow water. Journal of the Acoustical Society of America. 141:1501-1513.   10.1121/1.4976214   AbstractWebsite

This paper considers concurrent matched-field processing of data from multiple, spatially-separated acoustic arrays with application to towed-source data received on two bottom-moored horizontal line arrays from the SWellEx-96 shallow water experiment. Matched-field processors are derived for multiple arrays and multiple-snapshot data using maximum-likelihood estimates for unknown complex-valued source strengths and unknown error variances. Starting from a coherent processor where phase and amplitude is known between all arrays, likelihood expressions are derived for various assumptions on relative source spectral information (amplitude and phase at different frequencies) between arrays and from snapshot to snapshot. Processing the two arrays with a coherent-array processor (with inter-array amplitude and phase known) or with an incoherent-array processor (no inter-array spectral information) both yield improvements in localization over processing the arrays individually. The best results with this data set were obtained with a processor that exploits relative amplitude information but not relative phase between arrays. The localization performance improvement is retained when the multiple-array processors are applied to short arrays that individually yield poor performance. (C) 2017 Acoustical Society of America.

Riahi, N, Gerstoft P.  2017.  Using graph clustering to locate sources within a dense sensor array. Signal Processing. 132:110-120.   10.1016/j.sigpro.2016.10.001   AbstractWebsite

We develop a model-free technique to identify weak sources within dense sensor arrays using graph clustering No knowledge about the propagation medium is needed except that signal strengths decay to insignificant level! within a scale that is shorter than the aperture. We then reinterpret the spatial coherence matrix of a wave field as a matrix whose support is a connectivity matrix of a graph with sensors as vertices. In a dense network, well separated sources induce clusters in this graph. The geographic spread of these clusters can serve to localize the. sources. The support of the covariance matrix is estimated from limited-time data using a hypothesis test with robust phase-only coherence test statistic combined with a physical distance criterion. The latter criterion ensures graph sparsity and thus prevents clusters from forming by chance. We verify the approach and quantify its reliability on a simulated dataset. The method is then applied to data from a dense 5200 element geophone array that blanketed 7 km x 10 km of the city of Long Beach (CA). The analysis exposes a helicopter traversing the array and oil production facilities.

Gemba, KL, Hodgkiss WS, Gerstoft P.  2017.  Adaptive and compressive matched field processing. Journal of the Acoustical Society of America. 141:92-103.   10.1121/1.4973528   AbstractWebsite

Matched field processing is a generalized beamforming method that matches received array data to a dictionary of replica vectors in order to locate one or more sources. Its solution set is sparse since there are considerably fewer sources than replicas. Using compressive sensing (CS) implemented using basis pursuit, the matched field problem is reformulated as an underdetermined, convex optimization problem. CS estimates the unknown source amplitudes using the replica dictionary to best explain the data, subject to a row-sparsity constraint. This constraint selects the best matching replicas within the dictionary when using multiple observations and/or frequencies. For a single source, theory and simulations show that the performance of CS and the Bartlett processor are equivalent for any number of snapshots. Contrary to most adaptive processors, CS also can accommodate coherent sources. For a single and multiple incoherent sources, simulations indicate that CS offers modest localization performance improvement over the adaptive white noise constraint processor. SWellEx-96 experiment data results show comparable performance for both processors when localizing a weaker source in the presence of a stronger source. Moreover, CS often displays less ambiguity, demonstrating it is robust to data-replica mismatch. (C) 2017 Acoustical Society of America.

Li, J, Gerstoft P, Gao DZ, Li GF, Wang N.  2017.  Localizing scatterers from surf noise cross correlations. Journal of the Acoustical Society of America. 141:EL64-EL69.   10.1121/1.4974147   AbstractWebsite

The backscattered travel-time structure is obtained by cross-correlating air-acoustic ocean surf noise recorded on microphone pairs (separation similar to 2m) on the beach. The scatterer is a 20 cm radius Polyvinyl chloride pipe 2.5m landside of the microphone array. Arrivals corresponding to the time-difference (travel-time difference between two scatterer-receiver paths) and scattered (travel time for receiver-scatterer-receiver path) waves emerge in the cross-correlation functions in a backscattering configuration. Theoretically, only a microphone pair is needed to locate the scatterer using the time-difference and scattered travel times. Localization of the scatterer is demonstrated with the microphone array on the beach. (C) 2017 Acoustical Society of America