Volume 42, Issue 2, August 1967

Computer Recognition of Connected Speech
View Description Hide DescriptionA system for obtaining a phonemic transcription from a connected speech sample entered into the computer by a microphone and an analog‐to‐digital converter is described. A feature‐extraction program divides the speech utterance into segments approximately corresponding to phonemes, determines pitch periods of those segments where pitch analysis is appropriate, and computes a list of parameters for each segment. A classification program assigns a phoneme‐group label (vowellike segment, fricativelike segment, etc.) to each segment, determines whether a segment should be classified as a phoneme or whether it represents a phoneme boundary between two phonemes, and then assigns phoneme label to each segment that is not rejected as being a phoneme boundary. About 30 utterances of 1–2 sec duration were analyzed using the above programs on an interconnected IBM 7090‐PDP1 system. Correct identification of many vowel and consonantal phonemes was achieved for a single speaker using the same speech material that was used for developing the recognition procedure. The time for analysis of each utterance was about 40 times real time.

Time‐Domain Bandwidth‐Compression System
View Description Hide DescriptionA voice narrow‐band‐transmission system (VONAX) is described. The concepts upon which the system is based are discussed, and an experimental system is described. The system operates in a digital transmission mode at data rates of 9600, 4800, 2400, or 1200 bits/sec. In the present form, the experimental device is being used as a laboratory, research tool in the general area of speech‐processing research.

Excitation and Measurement of Acoustic Ion Waves
View Description Hide DescriptionMeasurements of the phase velocity and the attenuation constant of acoustic ion waves in weakly ionized gases in the frequency range from 0.1 to 6 MHz are described. The experiments were performed in continuous rf discharges having ion plasma frequencies of the order of 1 MHz. A beam of plane ion wares was excited and received by grids with diameters much greater than the wavelengths. At frequencies well below the ion plasma frequency, the phase velocity is frequency‐independent and approximately given by the Tonks‐Langmuir speed; the predicted dependence on ion mass is verified. The attenuation in this frequency range is primarily due to ion‐neutral collisions. At higher frequencies, the phase velocity increases slightly, whereas the attenuation per wavelength falls somewhat off. These results can be explained by assuming non‐Maxwellian velocity distribution of the ion gas. The measured phase velocity and attenuation were found to be independent of the geometrical dimensions of the grids, independent of the signal voltage applied to the excitation grid, and not influenced by drifts of the charged particles.

Elastic Constants of Filamentary Composites with Rectangular Symmetry
View Description Hide DescriptionDispersion relations for the propagation of elastic waves in filamentary periodic composites with rectangular symmetry are obtained for wavelengths that are long as compared with the intercomponent spacings. This yields averaged elastic constants of the material in terms of geometrical data and the elastic constants of the constituents. General averaging rules are established for all nine independent elastic constants. These rules are valid for any spatially varying elastic moduli and density with rectangular symmetry in the elementary cell. In the special case of a square lattice, the formulas are evaluated in the Wigner‐Seitz approximation. For two constituents, this gives explicit expressions for all six independent elastic constants.

Sound Propagation in Lamellar Composite Materials and Averaged Elastic Constants
View Description Hide DescriptionDispersion relations for the propagation of elastic waves in a lamellar periodic composite are obtained for wavelengths that are long as compared with the intercomponent spacings. This yields exact explicit expressions for the complete set of five independent, elastic constants of the material, in terms of geometrical data and the elastic constants of the constituents. The expressions are valid for any spatially varying elastic moduli and for density‐showing periodic structure and mirror symmetry within the elementary cell.

Ducted Acoustic Gravity Waves in a Nearly Isothermal Atmosphere
View Description Hide DescriptionThe propagation of acoustic gravity waves is analyzed in atmospheres that have vertical profiles of sound speed squared of the form , where , and both c ^{2}(∞) and Δc ^{2} are constant. Ducted wavedispersion relations are obtained from the requirements that the earth's surface act as a rigid boundary to the atmospheric oscillations and that the integral of the acoustic energy density over an infinite vertical column be finite. Ducted modes corresponding to the Lamb wave in an isothermal atmosphere and a surface wave in an incompressible medium are obtained for both signs in the above profiles. In addition, for the profile with the negative sign, an infinite number of closely spaced internal gravity and acoustic modes is shown to exist.

Echoes from Spherical Shells in Air
View Description Hide DescriptionCalculated results show that the elastic reaction caused by sound waves incident on a solid spherical shell in air can have a noticeable effect on the resultant echoes. The effect appears to be strongest for thin shells made of light material. For polyethylene, for example, the effect is found to exist for all thicknesses of the shell. It is suggested that the elastic response of an acoustic target might be used by bats in echolocation.

Numerical Solution for Transient Scattering from a Hard Surface of Arbitrary Shape—Retarded Potential Technique
View Description Hide DescriptionThe problem of transient scattering from a hard surface S reduces to the solution of an integral equation for the potential at S. To allow numerical solution,S is divided into K zones, on each of which the potential is assumed constant, and time is also divided into discrete intervals. These approximations lead to a set of K algebraic equations relating the present values of the potential to the retarded values and to the incident field. When the zoning is done properly, each equation involves only one present value, which can therefore be calculated explicitly from the retarded values. It is important to compute the coefficients in the Kequations accurately. This is accomplished by a numerical technique which takes into account the curvature of the zones and the variation in time retardation across zone. For scattering from a sphere, results agree very well with those obtained by separation of variables.

Axisymmetric Vibrations of a Solid Elastic Cylinder Encased in a Rigid Container
View Description Hide DescriptionThis paper describes the effects of end boundary conditions on the vibrational characteristics of solid propellent rocket motors. Previously, in the literature, solutions have been based on those obtained for an infinitely long cylinder. These solutions yield only certain sets of possible end boundary conditions for a finite cylinder, but not those considered here (i.e., fixed on all boundaries). The present approach consists of choosing a series of functions with unknown coefficients. Each term satisfies the governing differential equation and the boundary conditions on the axial displacement. The boundary conditions on the radial displacement are approximated by an orthogonalization procedure. This method yields an eigenvalue matrix the coefficients of which are transcendental functions of the frequency. The accuracy of the final solution depends on how well the radial displacement boundary conditions are satisfied. It is found that the procedure converges, and that sufficient accuracy is achieved, through the use of 20 terms in the series. The range of application of the simpler method, based on an infinite cylinder, is discussed by comparing results obtained from both methods.

Mobility of a Long Cylindrical Shell
View Description Hide DescriptionResults of an experimental and theoretical study of the mobility of a long, moderately thick‐walled, pipe‐like, cylindrical shell are reported. Experimental data for mobility are compared, over a wide frequency range not previously investigated, with analytical predictions based on thin‐shell theory. Existing solutions for free vibrations of thin shells are extended to predict vibration response to a point force that varies sinusoidally with time, based on a normal‐mode series solution. Practical convergence of the series is of special interest. The long, pipelike shell was chosen to make this convergence problem more severe since the resonances in some frequency ranges are very closely spaced. For most of the study, the ends of the experimental shell were free to move laterally, but the end cross section was restrained to maintain a circularshape; consequently, the shell was treated as free for the 1 = n or beam modes, and freely supported for the n = 2 or lobar modes. Some resonant frequencies are also included for the shell with completely free ends. Boundary conditions for the free‐free shell are discussed.

Vibration of A Rigid Circular Disk on an Infinite Isotropic Elastic Plate
View Description Hide DescriptionThe solution for a rigid circular disk vibrating vertically on an infinite elastic thick plate is investigated. Two cases are analyzed in which (a) the rigid disk is vibrating on an infinite plate that rests on a rigid immovable base, and (b) the rigid disk is vibrating on one surface of the plate while the other surface is kept free. The dual integral equations technique of Lebedev and Ufliand is utilized to solve the mixed boundary‐value problems. Numerical results are presented to illustrate the method.

Axisymmetric Torsional Wave Propagation in Circular Viscoelastic Rods
View Description Hide DescriptionA semi‐infinite viscoelastic circular cylindrical rod is subjected at its end surface to an arbitrary radial, axially symmetric tangential shear stress distribution, harmonic in time. Expressions are derived for the tangential and radial shear stress distributions, as well as for the tangential displacement. Near the forced end of the rod, the tangential displacement takes on a complicated form which consists of the superposition of a simple rotational oscillation of the undeformed cross section and motions deforming the cross section in its plane. The last type of motion permits infinitely many nodal circles. Owing to the large magnitude of the space attenuation coefficients, this second type of motion dies out rapidly and only the simple rotation, which is less attenuated, remains. Numerical computations were performed for two polymethyl methacrylate rods of 1 and 10 cm diam, subjected to a shear stress distribution varying periodically with time and cubic in the radial distance from the center. The results are presented in graphical form.

Estimation of Exponential Power Spectra
View Description Hide DescriptionMany naturally occurring random processes, such as ambient sea noise, have the property that the logarithm of their power spectra is approximately a linear function of log frequency. It is difficult to estimate a power spectrum with a steep negative slope due to the leakage of low‐frequency power into the higher‐frequency bands by the digital filter of the estimator. This spectral estimation problem can be simplified by assuming a two‐parameter model of the spectrum S (f). Let X (t) be a stationary Gaussian process with zero mean and , where θ and σ^{2} are unknown parameters. The total power of X (t) is σ^{2} and the mean frequency is θ^{−1}. These parameters are estimated by simple functions of the number of zero crossings of X (t) and the number of fixed level crossings in a given finite record. The estimator of ln S (f) is , where σ̂^{2} and θ̂ are the parameter estimates. For fixed σ^{2}, the large sample variances of θ̂ and lnŜ are derived and computed for , 1, and 2. Monte Carlo methods are used to estimate the medium sample properties of θ̂ and σ̂^{2}. The variance of lnŜ is of the order of T ^{−1}, whereas the variance of the standard digital estimator of lnS is of the order , where T is the record length.

Sea Lion Echo Ranging
View Description Hide DescriptionRecordings of a short series of sea lion pulses are analyzed. The approach used in the analysis is to infer echo‐ranging capabilities based upon the full utilization of the signal properties. The analysis has led to the hypothesis that, in addition to echo‐ranging pulses and possible communication signals, the sea lions—at least in some circumstances—generate and transmit signals that are used in the detection of target returns of echolocation pulses. In this framework, the actual echolocation pulse is the precursor pulse, and the main pulse is generated as part of the detection procedure. The resulting cross‐correlation detection approximates the optimum detection technique. Finally, it is concluded that, for the signals considered, range may be measured to within approximately ±4 cm and that, for range rates of interest—say, up to 5 m/sec—the signals provide essentially no range‐rate discrimination.

Auditory Sensitization
View Description Hide DescriptionIn three experiments, we have attempted to demonstrate the phenomenon of sensitization. We believe that we have succeeded. A brief high‐frequency tone was added to a low‐frequency carrier at various phases of the low frequency. Classical threshold was shifted up when the stimulus was added between 200° and 360°, and threshold was shifted down between 0° and 180°. This effect was greatest when the low frequency was at −5 dB sensation level (SL). At 20 dB SL the effect was partially obscured by masking.

Lateralization and Intensity Discrimination
View Description Hide DescriptionThis study reports the effects of various binaural intensity patterns on the discrimination of changes in the position and intensity of binaurally induced auditory images. The difference limen obtained for just‐noticeable shift of an auditory image is in most cases not significantly different from the difference limen observed in discrimination of a just‐noticeable change in the intensity of the signal. Only under low intensity levels is such a difference noted. Significant effects are observed as a function of the binaural intensity pattern employed. Also an increase intensity of the signal invariably produced a smaller just‐noticeable difference than a decrease in the intensity of the signal. A functional relationship is indicated between image movement and sensitivity to intensity change.

Detection and Localization: An Extension of the Theory of Signal Detectability
View Description Hide DescriptionThe theory of signal detectability is extended to detection and localization (DAL). We treat the following problem: on each trial there are k noisy observation intervals, and at most one of these contains a known signal; an observer makes one of k+1 responses, indicating “NO SIGNAL IS PRESENT” or “THE SIGNAL IS IN INTERVALj” (1⩽i⩽k). The observer is scored “correct” on a noise only trial if he reports “NO SIGNAL” he is scored “correct” on a signal‐plus‐noise trial if he correctly localizes the observation interval containing the signal. We derive the decision rules for maximizing percent correct, P(C); these are based on the likelihood ratio of the observation in each interval. Our analysis shows that the ideal DAL receiver consists of the sequential application of YES‐NO (YN) and forced‐choice (FC) decision‐makers. Under the assumption that monotonic transforms of the likelihood ratio are normally distributed, we obtain numerical results for P(C) as a function of k and the statistical separation of the noise and signal‐plus‐noise. Ideal performance on the DAL and FC problems is compared, and some multisignal cases are briefly considered.

Relation of Induced Tinnitus to Physical Characteristics of the Inducing Stimuli
View Description Hide DescriptionAfter observers were exposed to intense pure‐tone and broad‐band acoustic stimuli, their temporary threshold shifts were measured, and they were asked to match the pitch of any resulting tinnitus by manipulating the frequency of an adjustable low‐level pure tone in the opposite ear. It was found that both the frequency of tinnitus and the frequency of the tone used for the pitch match increased as the frequency of the traumatic stimulus increased, but maximum loss frequency and tinnitus frequency did not coincide. Although the observers were quite reliable in their judgments of tinnitus elicited by a stimulus, interobserver variability was considerable.

Temporal Summation for Critical Bandwidth Signals
View Description Hide DescriptionTemporal summation patterns were established for noise signals with critical‐bandwidth frequency spectrums centered at 1000 and 4000 cps. Critical‐bandwidth levels were calculated from group‐mean masked thresholds and converted to critical‐bandwidth noise signals. These measures were conducted for nine stimulus durations ranging from 10 to 1000 msec in ten normal‐hearing, and ten cochlear‐impaired subjects. Temporal summation functions for critical‐bandwidth levels and critical‐bandwidth signal thresholds were in good agreement except at 4000 cps in the cochlear‐impaired group. These subjects showed no change in absolute threshold for the critical‐bandwidth noise signals as a function of increasing stimulus duration time.

Beats of Mistuned Consonances
View Description Hide DescriptionTwo simultaneously sounding simple tones of M and N cps, respectively, with M:N slightly different from m:n (both small integral numbers) may give rise to a beat sensation with mN − nM beats per second. For the case m = 1, these beats are usually explained as resulting from harmonics created in the ear by distortion. Experiments indicated, however, that for m ≠ 1 secondary tones cannot be the main origin of these beats because the beat sensation does not disappear when bands of noise are introduced in the frequency ranges around possible harmonics and combination tones. Other experiments in which, for a constant lower toneM, the sound‐pressure level (SPL) of the higher toneN for most distinct beats was determined showed that this SPL decreases gradually for increasing frequency of N, irrespective of frequency ratio. Moreover, it was found that the beats for 200+601 cps manifest themselves in a weak tone sensation with a pitch shifting periodically in a sawtoothlike manner between about 610 and 730 cps. These experimental results suggest that, also for the case 1:n, the beats are not due to combination tones or aural harmonics but are related to periodic variations in the waveform of the overlapping vibration patterns along the basilar membrane, giving rise to corresponding variations in the time pattern of the nerve impulses.