Passive Acoustic Localization Research Articles

Artificial neural networks for simultaneous and independent range and depth discrimination in passive acoustic localization.

Two feedforward neural networks (NNs) with one hidden layer each were trained using a modified backpropagation algorithm to determine the position of an acoustic source in a waveguide. One network was trained to localize the source in depth while the other was trained independently to localize in range. The signal was preprocessed by decomposition along an orthogonal basis vector set in order to increase the robustness of the resulting trained network to uncertainties in the signal and environmental parameters. The output layer consisted of one unit for each possible range or depth of the source. The NNs were trained with a signal-to-noise ratio (S/N) of 50 dB and tested with patterns generated with S/Ns ranging from 50–0 dB. The performance of the NNs was compared with that of a conventional nearest neighbor processor. Evaluation of the processors was done in the context of an estimation problem, i.e., by measuring the standard deviation of the processors’ outputs. The NNs turned out to be less resistant to noise than the conventional processor, but were faster.

Source localization using the PE method

The most important information for source localization is oceanographic knowledge supplied to a powerful machine that numerically computes acoustic Green's functions rapidly and accurately. By this means the distorting effects of the oceanic medium and its boundaries can be removed, thereby rendering the ocean transparent. Then targets can be detected, localized, and classified at low S/N ratios as though they were moving in free space, leading to the result there is “no place to hide.” As an illustration of this approach, we discuss in detail the solution of the problem of passive narrow-band acoustic localization (an instance of the inverse source problem) following the pioneering work of Bucker [J. Acoust. Soc. Am. 59, 368–373 (1976)], in which beamforming was made obsolescent. Using the PESOGEN (Parabolic Equation SOlution GENerator) computer system [J. Acoust. Soc. Am. Suppl. 1 75, S26 (1984)] and an algorithm based on the generalized principle of reciprocity, we have demonstrated theoretically that targets can be accurately and reliably localized at long ranges and low S/N ratios using sparse configurations of sensors.

Passive acoustic location of bowhead whales in a population census off Point Barrow, Alaska.

A sonobuoy array placed in the nearshore lead was used for locating bowhead whale sounds to determine if whales migrated past census stations beyond visual range and were uncounted. Based on a sample of 182 whale sounds (over 48 h) from closest point of approach (CPA) distances out to more than 10 km, 68% originated beyond 2 km (CPA), where only 1% of the 242 whales were sighted. No whales were sighted beyond 3 km during this time, but 53% of the located sounds originated that far and beyond. Thirty-seven other bowhead sounds over 15 h were distributed out to 6 km. Two tracked whales moved at average speeds of 1.5 and 1.8 kn. Maximum location error was 1%-25% in a sector of 120 degrees X 5-10 km, depending upon bearing and range. Most whale sounds were low-frequency moans, trumpeting roars, and repetitive sequences (songs) with peak spectrum source level up to 189 dB re: 1 microPa, 1 m. Lack of correlations between numbers of sounds and sighted whales precluded using bowhead sounds to count individuals or even to extrapolate ratios of unseen to observed whales.

Information Derivable from the Doppler Polar Profile Spectrum

There has been considerable interest recently in the use of Doppler shifted signal information as a technique for passive acoustic localization. An investigation of this phenomenon using conventional techniques, however, rapidly becomes complicated by the large number of parameters and geometrical configurations that must be considered. In particular, it is especially difficult to determine the sensitivity of the technique to measurement errors. As an alternative, this paper treats the Doppler profile in terms of its inherent information content and indicates how such information may be utilized to obtain both qualitative and quantitative results. By invoking Fourier transform and sampling techniques, the limitations of Doppler localization are viewed in terms of conventional filter bandwidth and aliasing problems. Furthermore, measurement sensitivity is treated as a generalized S/N ratio enhancement problem. Utilizing the methods of this paper, a more generalized and complete understanding of the Doppler phenomena is achieved when compared to the specialized results of conventional methods.