Naval Electronic Systems Command Research Articles

Pulse synchronous analysis of speech using an auditory representation

Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA 02139 A procedure has been developed that automatically extracts pulse excitation information from the speech signal. This time‐domain algorithm operates on the transduction stage outputs of an auditory model developed by Seneff [this meeting]. Specifically, robust inflection points are determined from a waveform obtained by summing the channel outputs. This pulselike representation provides information about periodic and aperiodic excitation, and can be used to extract periodicities in voiced regions. Thus the information provided by this signal complements the spectral information provided by mean‐rate response outputs of the auditory model. Furthermore, this signal can be used to sample channel outputs pulse synchronously. This produces a spectral representation with sharper temporal onsets and offsets, and with an enhanced dynamic range in the frequency dimension. [Work supported by DARPA under Contract N00039‐85‐C‐0254, monitored through Naval Electronics Systems Command.]

A model for the transduction stage of auditory speech processing

A new model for the transduction stage of auditory processing is proposed. The model is applied independently to each of 40 critical band filter outputs, spanning the frequency range from 100–6400 Hz. The model consists of four components in cascade: an instantaneous nonlinear half‐wave rectifier, a short‐term adaptation component, a low‐pass filter, and, finally, a rapid adaptation component. Model outputs were compared with available auditory data in a number of different dimensions as follows: (1) onset response to tone bursts, (2) forward masking effects, (3) steady‐state behavior for vowel‐like stimuli, (4) incremental response characteristics, and (5) synchrony loss at high frequencies. Two distinct spectral representations are computed from the model outputs: an amplitude spectrum corresponding to mean‐rate response and a synchrony spectrum which makes use of temporal information to enhance spectral peaks. The model is currently being incorporated into a speaker‐independent continuous speech recognition system under development. [Work supported by DARPA under Contract N00039‐85‐C‐0254, monitored through Naval Electronics Systems Command.]

LIGHTWEIGHT BROADBAND HF COMMUNICATIONS ANTENNA (LWCA)

ABSTRACTExternal communications is a critical element in the U.S. Navy design and utilization of a ship's combat system. The communications antenna system is a key factor in attainment of reliable circuit performance and reduction of electromagnetic interference (EMI). To maintain pace with improved ship manufacturing techniques and construction materials, along with design efforts to reduce topside generated EMI/RFI effects, an improved antenna design must also evolve.With the ever increasing complexity in the integration of the topside environment, the RF aspects of the antenna designs must be augmented by detailed analysis of the operating environment and the mechanical design if the goals of reliability and quality performance are to be achieved.The Naval Electronic Systems Command has developed a new “Broadband HF Communications Antenna.” This paper traces the design evolution and describes the processes in determining current design deficiencies, the design objectives to correct these deficiencies and the results obtained.

Comparison of vertical line array data taken at two depths in the deep ocean

Vertical line array data were acquired in the Northeast Pacific using a 31‐element vertical line array placed with the center of the array first at the sound channel axis and second at a depth approximately equal to the conjugate depth of a cw source, which was towed at a nominal depth of 100 m for ranges up to 600 km from the array. These time series data have been transformed to the frequency domain and beamformed for frequencies near the cw source frequency to determine the signal arrival structure at each array depth. Preliminary results show that there is a strong persistent multipath arrival structure at both depths even at these long ranges. Although this vertical arrival structure exhibits fluctuations with range, three or four strong arrivals are present during most of the data. Qualitative agreement was found between the measured vertical arrival structure and that computed using range‐independent acoustic models. Values of array signal gain computed for both depths were very similar, showing an approximate 3.5‐dB degradation from that predicted theoretically. [Work supported by Naval Electronic Systems Command, Code 612.]

Sensitivity of two high resolution beamforming techniques to the number of line array elements and signal‐to‐noise ratio

Two high resolution beamformers, maximum entropy method (MEM) and Wagstaff‐Berrou broadband (WB2) [R. A. Wagstaff and J. Berrou, J. Acoust. Soc. Am. 75, 1133–1141 (1984)], were compared with each other and the standard FFT beamformer to determine performance degradation when the number of line array elements and the signal‐to‐noise ratio (SNR) were decreased. Experimental data from a multi‐tonal CW source recorded on an equally spaced horizontal line array were used to investigate the main beam broadening resulting from decreased SNR for SNR values ranging from 30 dB to near zero dB. Additional CW data from an equally spaced vertical line array were used to investigate the main beam broadening which resulted when the number of array elements was decreased successively from 31 to 16, 8, 4, and finally 2. Preliminary results indicate that the main beam broadening due to the decrease in array elements for both MEM and WB2 are similar and that both of these methods offer an order of magnitude decrease in the −3 dB beamwidth with respect to the standard FTT beamformer. [Work supported by Naval Electronic Systems Command, Code 612.]

Omnidirectional and directional sea surface backscattering strength measurements at low frequencies and low grazing angles

Limited experimental data exists on surface backscattering strengths at low acoustic frequencies (below 300 Hz) and low grazing angles (below 20°) for deep ocean environments. We report here preliminary results of such measurements taken during four days of a sea test in the North Atlantic in September 1980 (sea states varied from 2 to 3). Four sets of approximately 20 small explosive charges (5‐lb TNT‐equivalent; bubble frequency of 77 Hz) were dropped beneath a towed acoustic array. We measured the direct path surface reverberation by processing the first 3 s of data, for mean grazing angles down to 8°, prior to receipt of the first bottom bounce at 6 s past detonation. The data were averaged over six frequency bands of 20‐Hz width from 150 to 270 Hz and were reduced to scattering strengths versus mean grazing angle and frequency by using the NRL reverberation model to remove propagation effects and to compute the grazing angles versus time. Typical omnidirectional surface scattering strengths decrease from −55 to −70 dB for mean grazing angles from 20° to 8° with little (<5 dB) frequency dependence over the measured frequency range. These results are consistent with previously published measurements. Further, the data have been beamformed to examine the azimuthal directionality of the surface backscatter (sea direction varied from 27° to 110° relative to array heading), and preliminary results will be reported. [Data acquisition funded by Defense Mapping Agency; analyses supported by Naval Electronic Systems Command, Code 612.]

Prediction of ocean basin bottom reverberation using a range‐dependent reverberation model

The NRL model for underwater acoustic reverberation incorporates range dependent bathymetry and sound speed profiles. An extended ray theoretic approach for propagation estimates is combined with boundary scattering models to predict, among other measures, reverberation time histories. This paper presents model predictions of ocean basin and margin bottom reverberation for a monostatic source‐receiver geometry at a deep water site in the Northeast Pacific. The reverberation time histories are calculated for a number of radials which are selected to include prominent features such as seamounts, fracture zones, islands, and continental slopes. These time histories are azimuthally convolved with a receiver beam pattern to approximate the total basin reverberation response. The strong backscattered returns from both the continental shelf of North America and large fracture zones are of particular interest. Their strength can be 30 to 50 dB above the normal reverberation decay levels, making them a serious interference, even when located in the beam sidelobes. These predictions provide valuable information for source‐receiver design, signal selection, and pulse repetition rates to be used in future deep sea experiments. [Work supported by Naval Electronic Systems Command, Code 612.]

A comparison of acoustic propagation predictions for surface duct environments using ray‐theoretic and parabolic‐equation computer codes

An important type of under water acoustic environment involves a deep water SOFAR channel with a superimposed (winter) surface duct capable of trapping acoustic energy. In this paper we examine the effects predicted by a computer model of three ducted sound‐speed profiles on acoustic propagation at 300 Hz. The purpose of this study was to better understand the capabilities and limitations (both theoretical and numerical) of a ray theoretic model for predicting mid‐ and long‐range signal transmission loss (TL) in the winter North Atlantic. Several source and receiver configurations (combinations with one or both in and out of the duct) have been studied. To test validity, the ray theoretic calculations are compared to previously accepted results predicted by a propagation model based upon the parabolic equation. It is found that both model predictions agree qualitatively (TL measurements and acoustic field patterns are comparable), while any quantitative discrepancies are explained by the theoretical or numerical limitations of the different approaches. We conclude that when both the source and receiver are at least several wavelengths from the sea surface, ray theory can be quite accurate in its prediction for these ducted environments. [Work supported by Naval Electronic Systems Command, Code 612.]

Comparison of spatial coherence of bottom‐interacting sound for Abyssal Plain and Abyssal Hills sites

Acoustic reflectivity experiments were conducted over two different ocean bottom types in the northeast Pacific. Explosive sources were dropped at ranges up to 50 km from a vertical hydrophone array that was suspended near midwater depth. The experimental geometry allowed time isolation of the bottom‐interacting arrival. The bottom‐interacting arrival received by the array was processed to obtain spatial coherence for frequencies in the 20‐ to 250‐Hz band. The behavior of spatial coherence vs frequency, grazing angle, and sensor separation is distinctly different for the two sites. At the Abyssal Plain site, coherence is nearly 1.0 at grazing angles 10° to 25°, decreases with increasing grazing angle until about 60°, and then increases after 60°. At the Abyssal Hills site, coherence is low (about 0.3 for 280‐m sensor separation) and variable for 10° to 60°, and increases after 60°. The results are consistent with the combined effects of scattering from a rough interface and changing sensor geometry with grazing angle. [Work supported by NORDA and Naval Electronic Systems Command.]

The influence of sediment rigidity on interface wave excitation

Stoneley waves traveling along the interface between a fluid sediment and a solid substrate are a loss mechanism affecting low‐frequency sound propagation in the oceans [K. E. Hawker, J. Acoust. Soc. Am. 64, 548–555 (1978); 65, 682–686 (1979)]. Here we report a study of the effect of sediment rigidity (shear wave propagation) on the excitation of Stoneley waves. A numerical model [P. J. Vidmar, T. L. Foreman, J. Acoust. Soc. Am. 66, 1830–1835 (1979)] of the plane‐wave reflection coefficient, R, of the ocean bottom was used to study the effects of a finite sediment shear speed, Cs, and attenuation on the characteristic Stoneley wave peak in the bottom reflection loss, RL = ‐20 log 10 (|R|). For a 100‐m‐thick sediment layer with a shear wave attenuation of 15 dB/m/kHz, shear speeds of 100, 200, and 400 m/s were investigated. At 25, Hz Cs = 0 (fluid sediment) had the expected peak in RL; Cs = 100 m/s increased the height of the peak; Cs = 200 ms broadened the peak and shifted it to a higher grazing angle; and at Cs = 400 m/s (a reasonable value for Cs at the substrate interface) the peak is no longer visible in the generally increased RL. Frequencies of 50 and 100 Hz were also investigated. These results show that sediment rigidity greatly modifies the Stoneley wave‐loss process. [This work was sponsored by Naval Electronic Systems Command and Naval Ocean Research and Development Activity.]

Normal‐mode propagation in deep‐ocean sediment channels

E. L. Hamilton has discussed the likelihood of ducted acoustic transmission in deep‐ocean sediment channels, just below the water‐bottom interface. The normal‐mode solution is presented here; it yields eigenfunctions, the associated eigenvalue equation, and the minimum acoustic frequency that allows trapping of any specified mode. This frequency increases nearly linearly with mode number. Much of the acoustic energy propagated by any trapped mode travels through lossy bottom sediment. Accordingly, equations are derived for the attenuation coefficient of the nth mode. Numerical results are given for an “average” channel at a frequency of 100 Hz. [This work was sponsored by Naval Electronic Systems Command and Naval Ocean Research and Development Activity.]

A ray path analysis of sediment shear‐wave effects on bottom reflection loss

The effect of sediment shear‐wave propagation on bottom reflection loss (RL) from a typical deep sea sediment layer is analyzed by means of a ray path decomposition of the acoustic field. The purpose of this analysis is the development of a simple physical model of the sediment shear‐wave effects reported in a recent computational study of RL from a typical deep sea turbidite layer [P. J. Vidmar, J. Acoust. Soc. Am. Suppl. 1 66, S75 (1979)]. The ray path decomposition uses the interface reflection and transmission coefficients to obtain the amplitude of each shear (S) and compressional (P) wave. These coefficients are analyzed by means of an expansion in the parameter ε which is proportional to the relatively small S wave velocity of marine sediments. Ray paths with amplitudes up to order ε2 are identified and used to develop a qualitative ray path model of RL. P‐wave conversion at the sediment‐substrate interface is the dominant mechanism for sediment S‐wave excitation. To order ε2 RL is the result of the interference of three waves in the water whose amplitudes are proportional to ε0, ε1, and ε2. Comparison with RL obtained from a computational model shows that the ray path model correctly predicts the dependence of RL on frequency, S‐wave speed, S‐wave attenuation, P‐wave attenuation, and sediment thickness. [Work supported by Naval Ocean Research and Development Activity and Naval Electronic Systems Command.]

The effect of sediment rigidity on bottom reflection loss

The effect of sediment rigidity on bottom reflection loss, RL = −20 log10 (|R|) where R is the plane wave reflection coefficient, is studied using a computational model [P. J. Vidmar and T. L. Foreman, J. Acoust. Soc. Am. 65, S43 (1979)]. A single inhomogeneous turbidite layer overlying a homogeneous substrate is treated with an emphasis on low frequencies (10–200 Hz) and low grazing angles (0°–45°). We find that sediment rigidity can be neglected for thick (∼500 m) layers while it can produce significant increases in RL for thin layers (∼36 m). The frequency dependence of RL for thin layers has a low frequency regime characterized by large peaks (∼25 dB) recurring at short (∼2 Hz) intervals. At high frequencies there remains a persistent increase in RL (∼4 dB). A ray picture, based on an expansion of the reflection and transmission coefficient in the small sediment S wave speed, provides a basis for understanding these results. The most important mechanism for sediment S wave excitation is found to be compressional wave conversion at the sediment‐substrate interface. [Work supported by the Naval Ocean Research and Development Activity and the Naval Electronic Systems Command.]

Adaptive focusing: The likelihood ratio

A prerequisite for the implementation of the adaptive focusing detection scheme [A. J. Claus and F. M. Labianca, abstract F2, this session] is a temporally stable signal field distribution across the array aperture. Stability is characterized by a low‐rank signal covariance matrix. The case of absolute stability, that is, unity rank, occurs rarely in practice, but is of interest because its theoretical treatment can be carried out in closed form and lends insight to the choice of practical test statistics in less stable cases. The model is constructed under the assumption that linear filtering of the acoustic array data produces a vector time series in p‐dimensional unitary space U, where p is significantly smaller than the array dimension. Under H0 (noise only) the process in U is centered isotropic Gaussian, white of unit power. Under H1 (noise plus signal) the signal has constant but unknown ray direction in U, and the amplitude process along this ray is white (complex) Gaussian of specified power. The joint likelihood ratio for L observations in U is found in closed form, suitable for numerical calculation; it is a function of the eigenvalues of the p × p sample covariance matrix. The results of Monte Carlo simulation of the model, including cumulative distributions, ROC curves and performance curves are presented with emphasis on application to cases which might occur in practice. [This work has been supported in part by the Independent Research and Development Program of the Department of Defense and in part by the Naval Electronics Systems Command, Code ELEX‐320.]

Adaptive focusing: A spatial distortion‐adaptive receiver

Adaptive focusing is an array‐processing scheme in which the receiver is designed for adaptive detection of localized sources under conditions where the signal field distribution across the array aperture is distorted from a plane‐wave distribution, but is unknown a priori. The mechanisms responsible for the distortion, whether simple multipath or more complex propagation conditions, do not play important roles in the scheme. Rather, the emphasis is on estimation of the complex sensor responses using frequency domain processing and subsequent use of the resulting response vector to steer the array and thus maximize its output power for the signal sector of interest. The procedure exploits the spatial structure of the source field as it manifests itself in the eigenstructure of the sample covariance matrix. Whatever the physical mechanism causing the signal field distortion, the eigenstructure is generally such that estimation can be carried out in a unitary space of dimension significantly smaller than the number of sensors in the array. Clearly, temporal stability of the field distribution and signal‐to‐noise ratio are important parameters. Their influence on the detector has been studied and numerical results are presented. Monte Carlo simulation has shown that for threshold signals and under conditions for which a conventional beamformer is degraded by as much as 5 dB, adaptive focusing can perform within a fraction of 1 dB of an ideal beamformer for which the signal field was known a priori. [This work has been supported by the Naval Electronics Systems Command, Code ELEX‐320.]

An examination of multipath processes in a range dependent environment within the context of adiabatic mode theory

An extension of the work of Tindle and Guthrie [C. T. Tindle and K. M. Guthrie, J. Sound Vibration 34, 291–295 (1974)] allowing one to extract ray path information within the context of adiabatic mode theory has been carried out. The purpose of this extension is to examine the adiabatic solution with respect to the process of multipath conversion due to a sloping bottom to determine whether this process is contained within the adiabatic solution or if it is entangled in the coupling coefficients which are neglected within the adiabatic approximation. Through the application of Tindle and Guthrie's method to the description of acoustic propagation over a sloping bottom, it has been possible to show that the multipath structure of the field is described quite well by the adiabatic solution for bottom slope angles up to 10°. [Work supported by the Naval Ocean Research and Development Activity and the Naval Electronic Systems Command.]

Geometric sensitivity of frequency spreading due to acoustic scattering from wind driven water surfaces

Several recent studies have focused on the degree to which acoustic scattering from wind driven water surfaces exhibits frequency spreading which is asymmetric in any regular way. Ocean experiments, physical model studies and analytical models have so far not been entirely in agreement on this issue. Therefore a set of physical model experiments has been performed to investigate the sensitivity of frequency spreading asymmetries to variations in geometric parameters. These experiments were conducted using broadband probing signals in the model tank at Yale. Grazing angle, azimuth, projector and receiver beam pattern, acoustic frequency, and surface roughness were varied in a systematic way. The results of this series indicate that frequency spreading is in fact quite sensitive to a number of geometric parameters and that most current analytical models are inadequate to explain the detailed behavior of this phenomenon. [Work supported by NSF and Code 320, Naval Electronic Systems Command.]Several recent studies have focused on the degree to which acoustic scattering from wind driven water surfaces exhibits frequency spreading which is asymmetric in any regular way. Ocean experiments, physical model studies and analytical models have so far not been entirely in agreement on this issue. Therefore a set of physical model experiments has been performed to investigate the sensitivity of frequency spreading asymmetries to variations in geometric parameters. These experiments were conducted using broadband probing signals in the model tank at Yale. Grazing angle, azimuth, projector and receiver beam pattern, acoustic frequency, and surface roughness were varied in a systematic way. The results of this series indicate that frequency spreading is in fact quite sensitive to a number of geometric parameters and that most current analytical models are inadequate to explain the detailed behavior of this phenomenon. [Work supported by NSF and Code 320, Naval Electronic Systems Command.]

Experimental measurements with a large aperture PARRAY

A parametric acoustic receiving array (PARRAY) is characterized as a continuous, virtual end‐fire array of length equal to the separation between the pump and hydrophone. Therefore, the directivity index of a PARRAY is approximately equal to 10 log (4 L/λ) where L is the pump‐hydrophone separation and λ is the wavelength of the acoustic signal. A PARRAY with a pump‐hydrophone separation of 340 m has been installed in 45 m of water at the Lake Travis Test Station of Applied Research Laboratories, The University of Texas at Austin. Measurements over the frequency range from 40 to 800 Hz show spatial processing gains approximating the directivity index of the PARRAY. The geometry of the experiment will be described and data obtained with the 340 m PARRAY will be presented and discussed. [This research was supported by the Advanced Research Projects Agency of the Department of Defense and was monitored by the Naval Electronic Systems Command under Contract N00039‐76‐C‐0231.]

Transmitting SAW hydrophone

A transmitting hydrophone which remotely senses both absolute pressure and changes in pressure has been developed using a surface‐acoustic‐wave (SAW) oscillator. This oscillator is formed by placing a SAW delay line or resonator in the feedback loop of an amplifier. The pressure applied to the hydrophone is converted to a flexural loading of the piezoelectric plate which alters the surface‐wave velocity on the plate, consequently shifting the oscillator frequency. This frequency modulated signal is easily transmitted to a remote location for signal recovery. The piezoelectric plate thickness and length can be chosen to optimize the sensitivity of the SAW hydrophone. The surface‐wave element of the oscillator is a planar structure; therefore, it is simple and inexpensive to manufacture. Furthermore, the size and power requirements of the SAW hydrophone are small, typically less than 50 cm3 a and 300 mW. This paper presents performance data of a number of SAW hydrophones using SAW oscillators in the 40–100‐MHz range in conjunction with a versatile signal demodulation system. Extensive measurements of sensitivity and frequency response were made for both delay line and resonator hydrophones manufactured on LiNbO3 and S‐T cut quartz substrates and compared with measurements obtained from conventional hydrophones in the identical setup. [Work supported by Naval Electronics Systems Command Contract No. N00039‐78‐C‐0042.]

The influence of Biot's second dilatational wave on the reflection coefficient of ocean sediments

The theory of acoustic waves in a porous medium developed by Biot predicts two dilatational waves. Formulas for the reflection and transmission coefficients for normal incidence at an interface are available [J. Geertsma and D. C. Smit, Geophysics 26, 169–181 (1961)]. In the present work these formulas are extended to non‐normal incidence for the special case of an interface between water and porous sediments. Energy losses associated with Biot's second dilatational wave can be significant [R. D. Stoll, Geophysics 42, 715–725 0977)]. The theory will be used to predict bottom properties that can yield anomalous reflection coefficients and to propose experiments for the study of the second sound waves. [Work supported by Naval Electronic Systems Command.]