Index of content:
Volume 36, Issue 10, October 1964
- PROGRAM OF THE SIXTY‐EIGHTH MEETING OF THE ACOUSTICAL SOCIETY OF AMERICA
- Session A. Underwater Acoustics: Ocean Surface
- Contributed Papers
36(1964); http://dx.doi.org/10.1121/1.1919316View Description Hide Description
Experimental measurements have been made of both the peak and mean pressure amplitudes of acoustic pulses reflected from the oceansurface for a variety of sea states. The purpose was to investigate the influence of waveheight on the reflected signal. Pulsed sinusoidal signals were generated by a high‐power sound source bottom‐mounted in deep water and received on bottomed hydrophones at three ranges. The pulses were 100 msec in duration, at frequencies of 400, 750, 1000, and 1500 cps. Data were obtained at three grazing angles: 18°, 31°, and 85°. Statistics were calculated relating the variability of the reflected signal to grazing angle and waveheight. Results are given for waveheights ranging from 4 to 20 ft. In general, the mean pressure amplitude remained nearly constant with increasing waveheight at all three grazing angles. This is a continuation of work reported earlier [J. Acoust. Soc. Am. 35, 1892 (A) (1963)]. [Work supported by the Bureau of Ships, U. S. Navy.]
36(1964); http://dx.doi.org/10.1121/1.1939188View Description Hide Description
The scattering of high‐frequency underwater sound (λ≪waveheight) from the sea surface is described in terms of two mechanisms: (1) conventional scattering theory for a rough surface with normally distributed slopes appears to predict adequately the backscattering at incidence near normal; (2) a rather low population of near‐surface resonant bubbles would suffice to explain the behavior at angles near grazing. The two effects are used to interpret the experimental surface‐scattering data at high frequencies for all angles of incidence. [Hudson Labs. Columbia Univ. Informal Doc. No. 55.] [Work supported by the U. S. Office of Naval Research.]
36(1964); http://dx.doi.org/10.1121/1.1939189View Description Hide Description
A study was made at various sea‐state conditions of the coherence of an acoustic signal that has undergone one surface reflection. The source and receivers were bottom‐mounted in deep water near Bermuda. Two types of correlation measurements were made: (1) crosscorrelation of a surface‐reflected signal received at two hydrophones 500 ft apart, and (2) crosscorrelation of a surface‐reflected signal with a replica of the transmitted signal. The signal consisted of random noise for (1) and pseudorandom noise for (2). In both cases, the signal was limited to a 300‐cps band centered at 750 cps. Results indicate that the coherence decreases with increasing sea state in a linear manner. Normalized crosscorrelation coefficients obtained from the two hydrophones decrease from about 0.9 for a waveheight of 5 ft to 0.5 for a waveheight of 20 ft. Values obtained for the single‐hydrophone output and replica fall off at a much slower rate. [Work supported by the Bureau of Ships, U. S. Navy.]
36(1964); http://dx.doi.org/10.1121/1.1939190View Description Hide Description
A 12×4×4‐ft model tank is utilized to measure the underwater acousticbackscatter from a wind‐driven surface. A single transducer is used to transmit and receive a 0.1‐msec pulse at a repetition rate of 1000 pulses/sec. The carrier frequency is varied in the range from 200 kc/sec to 2 Mc/sec, which gives an acoustic wavelength/root‐mean‐square water‐waveheight ratio from 0.2 to 2.0. Observations of the relative energy return are made as a function of the incident angle for a fixed path length. A mathematical model is hypothesized that predicts the backscattered energy as a function of incident angle, transducer‐beam pattern, frequency, and surface statistics. Curves are presented that compare the observed data with the mathematical model. [This investigation supported by the Acoustics Program, U. S. Office of Naval Research, and the U. S. Naval Underwater Ordnance Station, Newport, R. I.]
36(1964); http://dx.doi.org/10.1121/1.1939191View Description Hide Description
The Kirchhoff approximation has been used in deriving the reverberant tail ps , which results when a short spherical pulse of strength F(t) backscatters from a rough pressure‐release surface. The amplitudes of the differential scattered pulselets producing ps depend on the displacement and slope of the surface elements at which they originate, the functional form of both F and F′, and also certain geometric weighting factors determined by the location of the elements. More‐specific insight is obtained from hs , the component due to scattering in the impulse response , of the “ocean‐surface filter.” The tail portion of hs may be expressed as a “ring average” of two terms, one of which involves the slope, the other the curvature of the surface. Near the specular region, the term involving slopes is dominant generally; later, the curvature term gains the upper hand. Specific calculations of for incident sine‐wave pings and explosive pulses reflecting from randomly rough surfaces are compared with recent experimental work of various investigators. [Work supported by the U. S. Office of Naval Research.]
36(1964); http://dx.doi.org/10.1121/1.1939192View Description Hide Description
Experimental curves are presented that describe the reflected sound field from a pressure‐release sinusoidal surface. The measurements were made in an 8‐ft‐diam, 7‐ft‐deep laboratory tank equipped with a dual‐transducer‐positioning system that is capable of positioning each transducer to any position in the tank with an accuracy of ±0.003 in. in any x, y, or z coordinate and ±0.1° in tilt and azimuth. The sinusoidal surface was machined in styrofoam sheets that were glued to a 4×4‐ft tray located near the bottom of the tank. The sinusoid was cut with an amplitude of 1.5 cm and a period of 4.5 cm. The amplitudes of the reflected pulses are presented as a function of the angle of reflection for several fixed grazing angles. Amplitude data are also presented as a function of frequency for several fixed grazing angles. Zero‐order or specular‐reflection data are presented for incident wavelengths varying from 1 to the amplitude of the sinusoidal surface. [Work supported by the U. S. Navy Bureau of Ships.]
- Session B. Speech Communication I
36(1964); http://dx.doi.org/10.1121/1.1939193View Description Hide Description
The function that maps the words of written English onto the corresponding words of spoken English is described. The simplest hypothesis is that this function F, defined on the symbols forming the letters of the alphabet, maps each letter onto a sound, and maps the sequence of letters L1L2 as . This hypothesis is false, since F is not always well‐defined in the sense that its values are not always unique and the equation does not always persist. On the basis of an exhaustive dictionary search, we have shown that: it is possible to define F in a context‐dependent manner such that its restriction to consonant strings is uniquely defined; with this new definition, the equation holds for consonant strings of the gramatically homogeneous one‐vowel string words of written English; the consonant strings in these words coincide with those uninfluenced by the rules of euphonic combination. [B. V. Bhimani, J. Acoust. Soc. Am. 35, 1912 (A) (1963); J. L. Dolby and H. L. Resnikoff, Language 40, No. 2, 167–196 (1964).] [Work supported by Lockheed Independent Research.]
36(1964); http://dx.doi.org/10.1121/1.1939194View Description Hide Description
It was hypothesized that results would differ for phoneme proportions taken from markedly different sources—for example: written versus oral material; highly familiar versus less familiar material; polysyllabic versus monosyllabic words; etc. Counts were made from phonemic transcriptions, using General American as a pronunciation guide, of the following samples: 500 most frequently occurring words of Thorndike‐Lorge; 500 words, each occurring less than once per million; all monosyllables from the first 500; a random sample of 100 of the first 500. These counts, in addition to the Dewey phoneme‐frequency count, and Tobias' transcription of the French, Carter, and Koenig data were evaluated by product‐moment correlations for raw data and coefficients of concordance for ranked data. Both types of indices yielded comparable magnitudes, varying between 0.85 and 0.97. The phoneme counts were correlated with acoustic and psychophysicalmeasures from several sources but no simple relations were forthcoming. [This investigation supported by the National Institute of Neurological Disease and Blindness, National Institutes of Health, U. S. Department of Health, Education, and Welfare, grant NB‐03162.]
36(1964); http://dx.doi.org/10.1121/1.1939195View Description Hide Description
Preliminary results are given on a comparative study of various objective talker‐recognition procedures, based on spectrographic analysis of 7 replicate utterances of each of 10 words by each of 10 different speakers. The spectrograms are quantized into 17 frequency channels and approximately 50 time channels. Different summarizations are applied to the spectrograms, including marginal energies, totalled across time, in each frequency channel; marginal energies for each time channel; and momentlike descriptions of energy distribution of the time margin. Various combinations of these summarizations were used as inputs to different multivariate distance measures, including (a) distance from unknown to a speaker centroid, using a metric based on a covariance matrix pooled over all speakers; (b) distances based on eigenvectors, using a classical discriminant‐analysis approach; (c) distances based on metrics, employing individual speaker covariance matrices. Percent correct identification varied from 22% (discriminant analysis, using one eigenvector of energy margin on time) to 97% [distance (a) applied to the frequency margins]. Frequency classification of energy is better than time classification; distance (a) is better than the others; certain words are much better than others.
36(1964); http://dx.doi.org/10.1121/1.1939196View Description Hide Description
Speech‐intelligibility scores as a function of noise level are studied for face‐to‐face, sound‐powered‐phone, and amplified‐speech‐ (earphone and loudspeaker) communication conditions. The speech‐interference level (SIL) for octaves of noise centered at 500, 1000, and 2000 cps (0.5/1/2) is used as the measure of noise level. By using this noise measure, much of the work in this field can be brought together and interpreted. It is noted that “noisy” and “very noisy” spaces are associated with SIL's such that “shouting” or “very loud” voice levels (or 95‐dB speech levels) are required for conversations at 1.5 or 3 ft, and this is the region where telephone conversations are judged to be “difficult” or “unsatisfactory.” All of these adverse noise conditions occur at the region where ear protection will aid intelligibility and at the boundary where ear protection should be used to protect against hearing losses. Where people must converse or communicate via some interior communication device (0.5/1/2), SIL's above 70 dB should be avoided. At levels greater than 90 dB (0.5/1/2) SIL, the wearing of hearing protection should be mandatory and every noiseproofing technique (except a noise shield for the microphone) should be employed. At levels above 100 dB SIL, every noiseproofing technique should be employed.
36(1964); http://dx.doi.org/10.1121/1.1939197View Description Hide Description
A number of CVC syllables containing the liquids /l/ and /r/, and produced by three American English talkers, have been analyzed using analysis‐by‐synthesis methods. The temporal variation of the lowest three formants was measured throughout the utterances. In the case of /l/, the frequencies of a zero and a pole in the vicinity of the third formant were also measured as a function of time. This pole‐zero‐pole cluster for /l/ varied considerably with talker and with vowel context, but values of F1 and F2 showed less variation with talker. Transitions examined in an F1‐F2 plane for both liquid consonants exhibited similar topological features for all talkers. Physiological interpretations of these data are discussed briefly.
36(1964); http://dx.doi.org/10.1121/1.1939198View Description Hide Description
By what mechanisms is a given phone produced in various phonetic contexts? Electromyographic studies of articulator action have been made to provide data bearing on this question. An earlier study attempted to specify the relation between observed myographic activity and tongue movement when a subject produced vowels in a “neutral” ([p]‐vowel‐[p]) context. In this study of the same subject, the four vowel nuclei [i], [u], [æ], and [ɔ] were spoken with all possible combinations of the lingual consonants [t] and [k], in CVC monosyllables. Muscle potentials were recorded from 5 sites spaced evenly along the dorsal surface of the tongue and one ventral site near the tongue tip. In many cases during vowels, voltage levels either increased or decreased as compared with levels seen in the neutral context. Some of these changes suggested that muscles can produce (a) increase in tongue movement in the same direction as that required for the following phone, or (b) decrease in movement if the following phone requires a contrary movement. Other features of the motor control of coarticulation are also illustrated and discussed. [This investigation supported in part by Public Health Service research grant No. DE 01774 from the National Institute of Dental Research, National Institutes of Health, U. S. Department of Health, Education, and Welfare.]
36(1964); http://dx.doi.org/10.1121/1.1939199View Description Hide Description
The purpose was to determine how quickly and accurately listeners could discriminate between pairs of synthetic stop consonants distributed along an acoustic continuum. The stimuli, synthesized on OVE II at the Royal Institute of Technology, Stockholm, were 13 consonant‐vowel syllables in which the second‐ and third‐formant transitions varied so as to produce /b,d,g/. Stimuli were presented in ABX format and listeners responded by pressing a button. Reaction time was measured from the onset of the third member of each stimulus triad to the onset of the response. The percentage of correct discriminations rose to a peak in the vicinity of the phoneme boundaries, while reaction times fell. These results contrast with those of a previous study on the identification of the same stimuli: in that study, consistency of identification fell in the vicinity of the phoneme boundaries, while reaction times rose. Some implications for an understanding of the processes of identification and discrimination are discussed. [This study supported in part by the National Science Foundation.]
36(1964); http://dx.doi.org/10.1121/1.1939200View Description Hide Description
The objective of this experiment is to compare the identification and discrimination of a series of rounded and unrounded vowels by Swedish and American English listeners. All stimuli were produced with the OVE II synthesizer at the Royal Institute of Technology in Stockholm. The unrounded vowels are in the phonemic system of both Swedish and American English, and encompass the three vowels /i/, /ɪ/, and /ɛ/. The rounded vowels encompass the Swedish vowels /i/, /y/, /u/, but are not in the phonemic system of American English. Each series consists of 13 vowels, for which the first three formant frequencies vary in approximately equal logarithmic steps through the indicated range. By discrimination tests showed that both series of vowels gave essentially identical discrimination data for both groups of listeners. Identification functions for the unrounded vowels were similar for the two groups of listeners, but American English listeners were less consistent than Swedish listeners in the identification of the rounded vowels. Implications of the results with regard to theories of speech perception are discussed. [This study supported in part by the National Science Foundation.]
36(1964); http://dx.doi.org/10.1121/1.1939201View Description Hide Description
As part of a long‐term study on the growth of patterning in children's language, a considerable body of spontaneous speech was collected from children in grades 1 through 5 and from adults. The speech was recorded with a pair of condensermicrophones on two tracks. Phonetic transcriptions were made using a 41‐symbol alphabet. Editing and normalization were shunned so that the statistical analysis could be used in studies of speech perception,learning, psycholinguistics, and linguistic structure. Phonetic transcriptions were analyzed by digital computer. The major data are (1) frequency distributions of singlets, pairs, and triplets, (2) mean phonetic “word” and “sentence” length, (3) the Markovian information in the sequential constraints of successive pairs and triplets, and (4) various distributions induced on the basic ones by linguistic rules, e.g., place and manner of articulation, stress, and intonation. Comparisons show the free speech of children and adults to be remarkably similar both in the distribution of units and in sequential structure. Implications of our findings for research in speech and psycholinguistics are discussed. [Work supported by the U. S. Department of Health, Education, and Welfare.]
- Session C. Geoacoustics I
36(1964); http://dx.doi.org/10.1121/1.1939202View Description Hide Description
Detailed interpretation of seismograms is dependent on the ability to decompose approximately the seismic disturbance into a set of events that have a simple theoretical basis. In the present paper, long‐range refraction propagation is considered in terms of geometrical ray theory. The various rays possible in a typical model differ only slightly in frequency content and arrival time, but differ substantially in horizontal (phase) velocity. Individual later‐arriving events can only rarely be identified on the conventional seismogram. Use of the frequency‐wavenumber description of an event for processing on a multichannel basis allows identification and separation of the desired arrivals, with suppression of other arrivals. Examples are given of the application of such processing to two nuclear events recorded on the cross array at Tonto Forest Observatory (TFO), Arizona. The outputs of the array are combined in such a manner as to enhance successively lower velocities. The resulting “seismograms” may be considered 3‐dimensional plots of amplitude versus velocity and arrival time. The PN and PG events are well‐defined. There are indications of P* for at least one of the two events. Scattered (low‐velocity) energy is negligible immediately after the arrival of PN , but is quite strong after PG . A likely scattering center is the sedimentary ridge north of TFO. The shear wave SG is also identified, and shown to be primarily transverse horizontal (SH). Theoretical computations are presented for the amplitudes of the various events, and the degree of similarity between theory and experiment is taken as a measure of our knowledge of the correct crustal structure.
36(1964); http://dx.doi.org/10.1121/1.1939203View Description Hide Description
An interferometric method has been employed in two ways to study the attenuation constants associated with compressional and shear waves. A “two‐faced” transmitter (at frequencies above 400 kc/sec) and two identical receivers (each equidistant from, and on opposite sides of, the transmitter), whose resonant frequency is well above that used to power the source, are the essential acoustic elements. The driving force is a pulsed sine wave of long enough duration to establish “steady‐state” conditions in the plate and yet avoid unwanted reflections. In the first arrangement (using one plate), the decreasing amplitude maxima of the transmission coefficient versus frequency are used to deduce the relation between attenuation and frequency. At frequencies (or thicknesses) where the transmission plot is exponential, two plates of different thicknesses are used to determine the attentuation constants directly. The reflection coefficients never enter the computations implicitly. The shear waves are “isolated” by properly choosing the incident angle. The various experimental limitations that have to be met are discussed briefly, together with some experimental results.
36(1964); http://dx.doi.org/10.1121/1.1939204View Description Hide Description
The 2‐dimensional problem of transient‐wave propagation in a layered, fluid, or solid half‐space with dipping interfaces is investigated to determine the effect of small tilt on reflected arrivals. The “ray‐theory” technique is used to obtain these expressions caused by an impulsive line load at the surface. The geometry of the medium is introduced into the analysis by rotating the coordinate axes in the physical space, which necessitates a change in the spatial transform variable. Numerical results are presented for some typical examples.
36(1964); http://dx.doi.org/10.1121/1.1939205View Description Hide Description
The design for a 30‐kbar pressure apparatus is presented. A sample holder containing a heater winding was designed to permit the measurement of the dilatational and the rotational wave velocities in solids. The experimental procedure for the apparatus is discussed. Dilatational and rotational wave velocities were measured, by the single‐pulse method, as the pressure and temperature were varied at pressures from 4.5 to 18 kbar and a temperature range from 22° to 1000°C. Samples investigated were Armco iron, 0.7% carbon steel, normal uranium, Burma jadeite, and Labradorite. A computer program was written to adjust the velocities for the change in the sample's dimensions due to the change in temperature and the change in pressure. The program calculates the elastic moduli of the sample and fits a linear or quadratic curve to the points (linear or quadratic in pressure and temperature).
36(1964); http://dx.doi.org/10.1121/1.1939206View Description Hide Description
A simple analysis shows that, when a transient plane wave in an elastic medium is reflected at a plane boundary with a lossy medium, the transient waveform of the reflection is affected by the loss parameters of the second medium. If the attenuation in the second medium is small and if the ρc products of the two media are matched, then the reflected waveform is the convolution of the incident waveform with the integral of the Fourier transform of attenuation as a function of frequency. Thus, attenuation for a lossy solid or liquid can be obtained by this external‐pulse technique. Where attenuation is some simple function of frequency, its Fourier transform is some recognized generalized function. Sample waveforms have been observed, using airborne sound in specially prepared tubes.