A neural network approach to source localization

Steinberg, Ben Z.; Beran, Mark J.; Chin, Steven H.; Howard, James H.

doi:10.1121/1.401635

Cited by 41 publications

(18 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning in ocean acoustics was conducted using theory and computational resources in the 1990s. [4][5][6][7][8][9] A recent example of machine learning in ocean acoustics is the application of nonlinear regression 10 to source localization. In addition, data-driven linear cross correlation methods were used to localize sources of opportunity in shallow 11,12 and deep 13 water.…”

Section: Introductionmentioning

confidence: 99%

Ship localization in Santa Barbara Channel using machine learning classifiers

Niu

Ozanich

Gerstoft

2017

The Journal of the Acoustical Society of America

119

View full text Add to dashboard Cite

Machine learning classifiers are shown to outperform conventional matched field processing for a deep water (600 m 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 10 km range whereas the conventional matched field processing fails at about 4 km range without accurate environmental information.

show abstract

Section: Introductionmentioning

confidence: 99%

Ship localization in Santa Barbara Channel using machine learning classifiers

Niu

Ozanich

Gerstoft

2017

The Journal of the Acoustical Society of America

119

View full text Add to dashboard Cite

show abstract

“…Back in the early nineties and in the first decade of the current century, works such as Refs. [ 40 , 55 , 56 ] proposed the use of neural network techniques in this area. However, an evaluation of realistic and extensive datasets was not viable at this time, and the proposals were somewhat limited in scope.…”

Section: State Of the Artmentioning

confidence: 99%

Towards End-to-End Acoustic Localization Using Deep Learning: From Audio Signals to Source Position Coordinates

Vera-Diaz

Pizarro

Macías-Guarasa

2018

Sensors

View full text Add to dashboard Cite

This paper presents a novel approach for indoor acoustic source localization using microphone arrays, based on a Convolutional Neural Network (CNN). In the proposed solution, the CNN is designed to directly estimate the three-dimensional position of a single acoustic source using the raw audio signal as the input information and avoiding the use of hand-crafted audio features. Given the limited amount of available localization data, we propose, in this paper, a training strategy based on two steps. We first train our network using semi-synthetic data generated from close talk speech recordings. We simulate the time delays and distortion suffered in the signal that propagate from the source to the array of microphones. We then fine tune this network using a small amount of real data. Our experimental results, evaluated on a publicly available dataset recorded in a real room, show that this approach is able to produce networks that significantly improve existing localization methods based on SRP-PHAT strategies and also those presented in very recent proposals based on Convolutional Recurrent Neural Networks (CRNN). In addition, our experiments show that the performance of our CNN method does not show a relevant dependency on the speaker’s gender, nor on the size of the signal window being used.

show abstract

“…Very rudimentary algorithms have been discussed for simple point source, free field geometries in terms of training networks. [92] The issue of extracting weak signals from noise by training a network using hign SNR data has also been approached. [74] It has long been recognized that the acoustic propagation in the ocean is a stochastic proceSSj nevertheless, MFP algorithms formulated to date have not incorporated this directly into their design.…”

Section: Other Mfp Algorithmsmentioning

confidence: 99%

Matched Field Processing in Ocean Acoustics

Baggeroer

Kuperman

1993

Acoustic Signal Processing for Ocean Exploration

View full text Add to dashboard Cite

ABSTRACT. Matched field processing (MFP) is a generalized beam forming method which uses the spatial complexities of acoustic fields in an ocean wavegnide to localize sources in range, depth and azimuth or to infer parameters of the waveguide it.self. It has experimentally localized sources with accuracies exceeding the Rayleigh limit for depth and the Fresnel limit for range by two orders of magnitude. MFP exploits the coherence of the modefmultipath structure and it is especially effective at low frequencies where the ocean supports coherent propagation over very loug ranges. This contrasts with plane wave based models which are degraded by modal and lUultipath phenomena and are generally ineffective when waveguide phenomena are important. MFP is a spatial matched fLlter where the correlation signal, or replica, is determined by the Green's function of the waveguide. It can have either conventional or adaptive formulations and it has been implemented with an a.-;sortment of both narrowband and wideband signal models. All involve some form of correlation betwcen the replicas deri ved from the wave equation and the data measured at an array of sensors. Since the replica gcnerally has a complicated dependence upon the source location and environmental parameters, the wave eqnation mllst be solved over this parameter space. One can view MFP as an inverse problem where one attempts to invert for these dependencies over the parameter space of the source and the environment. There is cUlTently a large literature discussing many theoretical aspects of MFP and this is supported by numerous simulations; several experiments acquiring data for MFP now have been conducted in several ocean environments and these have demonstrated both its capabilities and some of its limitations. Consequently, there is a modest understanding of both the theory and the experimental capabilities of MFP. This article provides an oVl!fvicw of both.

show abstract

A neural network approach to source localization

Cited by 41 publications

References 0 publications

Ship localization in Santa Barbara Channel using machine learning classifiers

Ship localization in Santa Barbara Channel using machine learning classifiers

Towards End-to-End Acoustic Localization Using Deep Learning: From Audio Signals to Source Position Coordinates

Matched Field Processing in Ocean Acoustics

Contact Info

Product

Resources

About