Volume 22, Issue 2, March 1950
Index of content:
22(1950); http://dx.doi.org/10.1121/1.1906605View Description Hide Description
The intelligibility of the speech received over a communication system is usually expressed in terms of one or another measure such as the vowel or the consonant articulation, the average speech sound articulation, the syllable articulation, the word articulation, or the sentence intelligibility. The present paper establishes relationships among several of these measures and the articulation index. Relationships based upon statistical considerations are compared with the results of observations. Functions are developed which permit the calculation of articulation index and hence of articulation for communication systems which include a wide variety of response versus frequency characteristics and of noise conditions, as well as several special types of distortion. Although the treatment is predominantly empirical, the functions and processes are closely related to various fundamental properties of speech and hearing. Four principal series of articulation tests are cited in detail, some of which have been described in published articles by various persons. The response and the noise, if any, are given for each of these cases and the observed articulations are compared with values calculated by the method here presented. The application of the computational method to the perception of speech by deafened persons is reserved for a subsequent paper. A “Foreword” to the present paper describes the historical importance of articulation tests in the Bell TelephoneSystem.
22(1950); http://dx.doi.org/10.1121/1.1906583View Description Hide Description
The physiological motions involved in speaking can be indicated to the eye or to the ear. For the eye suitably chosen symbols may be written to indicate the physiological positions assumed informing each sound; for the ear synthetic sounds may be produced by motions in a mechanism built to simulate the speechorgans. The degree of phonetic success may be estimated in the case of the visible symbols by listening to sounds formed when the indicated physiological processes are carried out, and in the case of the speech‐simulating mechanism by comparing the synthetic speech produced to normally spoken speech. Significant advances along both the visual and the aural lines are described from earliest times down to the present.
Wolfgang von Kempelen produced the first speaking machine worthy of the name around 1780. This paper gives his background, a description of the apparatus he built, and a discussion of the methods used in producing the various sounds, fitting his work into the over‐all picture of speech‐imitating devices from the speaking of idols of ancient times down to the automatic electrical reconstructing of speech in the vocoder. For portraying to the eye the physiological characteristics of speech there are discussed the more outstanding methods from claimed symbolic alphabets of ancient languages down to the recent spectrographic visible speech.
22(1950); http://dx.doi.org/10.1121/1.1906584View Description Hide Description
This paper concerns the effects of interrupting speech waves—turning them on and off intermittently or masking them with intermittent noise—upon their intelligibility. The effects were studied with various rates of interruption and with the speech left undisturbed various percentages of the time. Tests were conducted (1) with speech turned on and off in quiet, (2) with continuous speech masked by interrupted white noise, and (3) with speech and noise interrupted alternately, the speech wave being turned on as the noise wave was turned off, and vice versa.
(1) When the speech wave is turned on and off infrequently, the percentage of the message that is missed is approximately the same as the percentage of time the speech is off. When the interruptions are periodic and occur more often than 10,000 times per second, the interruptions do not interfere with the reception of the message. In the quiet it is easy to understand conversational speech so long as the interruptions occur more than 10 times per second.
(2) When continuous speech waves are masked by noise that is interrupted more than 200 times per second, intelligibility is independent of the interruption frequency and of the percentage of time the noise is on, provided the ratio of average speech power to average noise power is held constant. Interrupted masking noise impairs intelligibility least if the frequency of interruption is about 15 per second.
(3) When interrupted speech and interrupted noise alternate at frequencies below 10 alternations per second, the noise does not impair intelligibility. At higher frequencies of alternation the temporal spread of masking becomes appreciable.
The general features of these results are approximately the same whether the interruptions occur periodically or at random.
22(1950); http://dx.doi.org/10.1121/1.1906585View Description Hide Description
Groups of 23 males read 12 test phrases in each of eight rooms. The rooms represented two sizes, shapes, and reverberation times. Microphones led to two meters that registered vocal intensity and, in one instance, duration of the phrases. Each set of measurements was treated by analysis of variance. Both rate and intensity of reading were affected by the size and reverberation time of the room and not by the shape. Rate was significantly slower in the larger and the less reverberant rooms. Apparently vocal intensity was greater in the smaller and less reverberant rooms; and readers consistently increased their intensity as they read the 12 phrases in the less reverberant rooms. Differences in this regard, occasioned by the sound treatment of the rooms, were highly significant.
22(1950); http://dx.doi.org/10.1121/1.1906586View Description Hide Description
In the past loudnesspatterns have been based on the masking effect of one sound on another. For complex sounds having distributed energy spectrums this method appears to be valid. For sounds with single frequency components the method is thought to be in error due to the formation of beats and modulation products between the primary tone for which a pattern is desired and the probe tone which is used to determine the pattern details. To avoid these difficulties in the present tests, the probe tone was presented after the primary tone was turned off. The resulting residual masking patterns differ in a number of important respects from patterns based on the simultaneous masking procedure.
A comparison between the loudness of a primary tone, as evidenced by the magnitude of its residual masking pattern, with the results of loudness judgment tests was made. This was done by replacing the physical scales of pressure level and frequency by the subjective scales of loudness and position. A reasonably good check of computed and measured loudness values was obtained. Patterns for a 1000‐cycle tone were measured to show how the loudness of the standard reference tone is distributed and how this distribution changes as the level of the tone is increased.
22(1950); http://dx.doi.org/10.1121/1.1906587View Description Hide Description
In investigations of binaural hearing in which the interaural phase difference has been studied, either (1) the stimuli have been pure tones, in which case there is just one interaural phase difference, or (2) no attempt has been made to vary separately the several or many phase differences that exist, one for each frequency component, when both ears are stimulated by complex waves. In the present observations, we have studied the effect of varying the interaural phase relations in the simplest of complex waves: two superposed sinusoids of equal amplitudes. A two‐component tone was presented binaurally, and the interaural phase difference of one of the components was switched alternately from 0 to 180 degrees. The effect of the phase reversal upon the listener's subjective experience and the frequency dependence of the effect are described.
22(1950); http://dx.doi.org/10.1121/1.1906588View Description Hide Description
The binaural masked threshold for speech depends upon the relation between the interaural phase angles of the speech and those of the noise. When these phase angles are the same, the threshold is high, and both speech and noise appear to be in the same place. When the interaural phase angle of the speech is reversed relative to that of the noise, the threshold is low and the speech and noise appear to be in different places. These relations have been clearly demonstrated with earphones, and they suggest that in free‐field listening the threshold of intelligibility might be affected by the relative locations of the sources of speech and of masking noise. It was found that when the azimuths of the sources of speech and of noise are changed relative to each other, the threshold of intelligibility changes by small but consistent amounts. When the sources are close together, the threshold is high; when the sources are far apart, the threshold is reduced.
Although this relation is partially confounded by the effect of azimuth on the sound pressure levels at the ears, the factor of localization appears to play a significant role, especially when two ears are used and when the head is allowed to move. In order that the hard‐of‐hearing may take advantage of these effects, they must have a hearing aid with two separate microphones mounted near the ears and connected each to a separate earphone.
22(1950); http://dx.doi.org/10.1121/1.1906589View Description Hide Description
In the recent past a program was initiated to survey vehicle, traffic, and industrial noise in the Chicago area. The phase on noise of vehicles has been completed. The investigation included street, elevated, and subway cars; diesel, steam, and electric trains; and motor buses,trucks, and automobiles.
Measurements were made of over‐all and octave band levels. Inside of the vehicles, flat network over‐all levels ranged from 85 db in a new “L” car to 95 db in subway cars. The readings in the 400–800 c.p.s. band ranged from 68 db in an automobile to 91 db in subway cars. Outside of and close to vehicles, the flat network over‐all levels ranged from 78 db for automobiles to 94 db for subway trains. Observations in the 400–800 c.p.s. band ranged from 66 db for automobiles to 87 db for subway trains.
Variations in the over‐all levels inside of vehicles ranged from ±1 to ±5 db. Variations the 400–800 c.p.s. band ranged from ±2 to ±5 db. Outside of the vehicles, variations ranged from ±2 to ±6 db in both the over‐all and 400–800 c.p.s. band levels.
22(1950); http://dx.doi.org/10.1121/1.1906590View Description Hide Description
Direct‐radiator loudspeakers are often mounted with the back of the diaphragm working into a completely enclosed space. Conventional theory states that when the maximum linear dimension of such an enclosure is small compared with the wave‐length, the pressure is uniform throughout, and the acoustical impedance presented to the loudspeaker is ‐j/ω(V/ρc 2), where V is the enclosed volume. Although it has not been clearly established how small an enclosure must be before it is “small compared with the wave‐length,” the foregoing expression is generally used, at low audiofrequencies, to calculate the acoustical impedance of closed loudspeaker housings.
It is shown here that while the acoustical impedance of a closed rectangular housing is capacitive at very low frequencies, it passes through zero as the frequency increases and becomes that of an inertance as the frequency of the first normal mode is approached. For a typical housing 11 in. × 22 in. × 22 in., the frequency at which the impedance presented to a very small speaker passes through zero is in the vicinity of 70 c.p.s.; at this frequency, the maximum linear dimension of the enclosure is less than one‐seventh of the wave‐length.
These results are obtained by using the methods given by Morse for determining the pressure distribution in a room. A point‐source loudspeaker is assumed and the pressure at the source is calculated as the summation of the pressures due to the normal modes of the enclosure, losses being neglected. Measurements of the pressure within the enclosure support this analysis.
From measurements of the pressure distribution over the surface of the loudspeaker diaphragm, it may be deduced that the magnitude of the acoustical impedance which the enclosure presents to the loudspeaker diaphragm and the frequency at which the impedance becomes zero depend upon the dimensions of the loudspeaker diaphragm as well as the dimensions of the enclosure.
22(1950); http://dx.doi.org/10.1121/1.1906591View Description Hide Description
In his theory of streaming caused by sound waves, Eckart shows that time independent streams necessarily follow as part of the solution of the complete wave equatoin, taking into account viscosity and second‐order terms. His treatment is mainly valid for liquids and it proves that the driving force of the streams is proportional to frequency squared. The effect, therefore, is especially important in the ultrasonic region (crystal winds). However, he suggests that slow streams might also be carried in air at audio frequencies.
Studies of acoustical streaming phenomena around orifices have been made by the use of smoke particles in a 3‐in. diameter circular tube. These studies covered a range of orifices from thicknesses of 0.5 mm to 19 mm and diameters of 3.5 mm to 20 mm. The frequency lay between 150 to 1000 c.p.s. Velocities in the orifice cover the range of 0 to 700 cm/sec.
Close studies of the flow patterns have disclosed that there exist four definite regions of flow as the particle velocity in the orifice is increased. These regions have been represented by “phase diagrams.” Photographs of the various flow patterns in each region of the “phase diagram” have been taken for a number of orifices. Under each observed condition, the acoustic impedance of the orifice is determined by a conventional standing‐wave measurement in the tube.
It is shown that the nonlinear properties of the acoustic impedance of an orifice is closely connected with the circulation effects. Quantitative check in one of the circulation regions and a good qualitative over‐all agreement indicate that the nonlinear properties of the impedance is due to the interaction between the sound field and the circulatory effects.
22(1950); http://dx.doi.org/10.1121/1.1906592View Description Hide Description
Solutions for compressional waves in an isotropic rod of rectangular cross section are formulated from the general elasticequations of motion. Two general modes of propagation are constructed which satisfy the boundary conditions for the free vibrations of a long rod for which the width 2d is considerably greater than the thickness 2a. The thickness mode is shown to have an infinite phase velocity (cut‐off frequency) at that frequency for which the half wave‐length in an infinitely wide rod would be equal to 2d. The phase velocity of both of these propagation modes is shown to approach that of Rayleigh surface waves when the wave‐lengths are much smaller than the lateral dimensions. The calculated thickness mode cut‐off frequencies are compared with measurements reported earlier and the agreement is shown to be good. A comparison between the predicted and measured phase velocity dispersion is made for various values of a/d. The agreement is found to be satisfactory as long as d is several times greater than a.
22(1950); http://dx.doi.org/10.1121/1.1906593View Description Hide Description
Equations have been derived to give the torsional vibrational frequencies of a system consisting of two identical cylinders joined axially by means of a third cylinder. Such a system corresponds satisfactorily to the experimental set‐up frequently used to determine the modulus of rigidity of a sample. The relation of the frequency spectrum to the various parameters of the system is discussed in detail. Two sets of experimental frequency data are compared with the results of the theoretical equations.
22(1950); http://dx.doi.org/10.1121/1.1906594View Description Hide Description
Theoretical analysis predicts the occurrence of phase shifts when harmonic waves are reflected from interfaces of higher soundvelocity (such as the sea bed in underwater sound propagation) at angles exceeding the critical angle of total reflection. Since the phase shift depends only upon the acoustic parameters of the media forming the interface and upon the angle of incidence and is independent of the frequency of the incident wave train, one would expect pulses of arbitrary shape to be subjected to distortion upon reflection.
The expected shape of the pressurewave reflected from a semi‐solid bottom is derived for the specific case of exponentially decaying shock waves produced in underwater explosions, and the theoretical predictions are found to agree well with the observed pressure‐time curves of the first and successively higher order reflections. The theoretical analysis assumes acoustic plane waves incident upon an interface between two fluid media. Experimental records were obtained at relatively great distances from the charge so that finite amplitude effects were negligible and the radius of curvature of the wave front quite large.
22(1950); http://dx.doi.org/10.1121/1.1906595View Description Hide Description
An ultrasonicinterferometer has been developed using an air‐liquid surface as a reflector. Because of certain simplifications that occur, it is found advisable to recast the mathematical interpretation of the results along different lines than previously. The effect of a loaded piezoelectric crystal at resonance is taken as a variable resistance, and the measured values of this resistance are interpreted to calculate absorption in the liquid.
22(1950); http://dx.doi.org/10.1121/1.1906596View Description Hide Description
An interferometer has been developed which makes use of the free surface of a liquid as a reflector. The emitter is an optically plane quartz crystal which is held perfectly horizontal at the bottom of a vertical liquid column. Changes in the column height produce the effects which are used to measure absorption in the liquid. A sensitive method of adding liquid makes the instrument usable at frequencies well above six megacycles. The leveling is sensitive and positive.
22(1950); http://dx.doi.org/10.1121/1.1906597View Description Hide Description
The ultrasonicsounds emitted by bats have been analyzed with a system sensitive to frequencies from 1 to 150 kc. These sounds are used by the bats to detect obstacles by means of their echoes; and they consist of pulses about two milliseconds in duration. Most of the measurements were made with the common little brown bat, Myotis l. lucifugus; with this species the sound pressure at 40 to 55 kc, measured 5 to 10 cm from the animal's mouth, averaged 60 dynes/cm2 (109 db on the conventional scale of sound pressure levels). The highest recorded intensity was 173 dynes/cm2 (119 db). The frequency of the ultrasonicsound falls during each pulse by about one octave; the average frequency at the peak amplitude was 47.8 kc, while the average at the beginning of the pulse was 77.9 kc, and at the end 39.1 kc. Low frequency waves (about 10 kc) accompany the pulse, but their amplitude is a very small fraction (1/100 to 1/1000 or less) of the peak sound pressure at ultrasonic frequencies. The envelope form is variable; the emission is directional with most of the energy concentrated into the forward direction; and the pulses are commonly repeated at rates of 20 to 30 per second.
22(1950); http://dx.doi.org/10.1121/1.1906598View Description Hide Description
The application of electronic facilities to the control of sound in the legitimate theater has been subject of a long term research project subsidized by the Rockefeller Foundation, the Research Corporation, and the Stevens Research Foundation. Complete control of the auditory component of the show was achieved immediately prior to the war. A modular soundcontrol system has been developed, capable of filling all needs of the legitimate theater and specialized requirements of other entertainment forms.
22(1950); http://dx.doi.org/10.1121/1.1906599View Description Hide Description
The measurement of the acoustic absorption coefficient by a steady state method was carried out at frequencies of 9, 20, and 30 kc for seven different materials. This involved the construction of a sound chamber with facilities for creating a diffuse sound field and a sample area where materials could be mounted. The average intensity in the chamber was measured with the sample area covered with the material under test and the results compared with similar measurements when the area was covered successively with a material of negligible absorption and when it was open to the air outside. The expression giving the absorption coefficient in terms of these three relative intensity readings is derived.
22(1950); http://dx.doi.org/10.1121/1.1906600View Description Hide Description
A method is described for utilizing normal absorption coefficient or acoustic impedance measurements to predict reverberant sound absorption coefficients. The average of coefficients for the six standard frequencies determined from acoustic impedance measurements agrees closely with the average reverberant coefficient, for cases where the material may be said to obey the normal impedance assumption. The normal absorption coefficients of some 26 different acoustic materials were measured at 512 c.p.s Using the method given in the paper, the predicted reverberant coefficient deviated from the measured reverberant coefficient by 0.05 or less for 18 materials, while in only 3 cases were the deviations greater than 0.10. The method should be particularly applicable to the problem of acceptance testing of installed acoustic materials.
In the theoretical development, best agreement with experiment was obtained by introducing a new kind of reverberant statistics which associates with each wave packet in a random field a scalar quantity equal to the square of the absolute value of the sound pressure in each packet, instead of the customary energy flow treatment. Also, it was found necessary to carry out the analysis using a concept of equivalent real impedance to replace the usual complex impedance.
22(1950); http://dx.doi.org/10.1121/1.1906601View Description Hide Description
The transmission of reverberant sound through a double wall, which consists of two identical single walls coupled by an airspace, is investigated both theoretically and experimentally. A theory is developed which gives good agreement with experiment. In order to compute the transmission loss of a double wall, it is necessary to know the impedance Zw of the single wall. Zw was determined from experiments conducted on the single wall and includes the effects of mass, dissipation, and flexural motion. The treatment shows that it is impossible to get a large improvement in transmission loss for a double wall relative to a single wall under reverberant sound field conditions if the single wall is considered to have only mass reactance. In addition, the customary normal incidence theory is totally inadequate in explaining the behavior of a double wall in a reverberant sound field.
For double walls having air‐coupling only, very shallow airspaces can produce appreciable increases in transmission loss over a single wall. An absorbent material when inserted in the airspace produces large improvements only when the mass of the walls is relatively light and has but little effect for heavy walls. Honeycomb or other non‐absorbent cellular structures having no cell walls in a direction normal to the wall faces do not result in an increase in transmission loss. Air‐coupled walls having no solid sound conducting paths between individual septa are extremely effective sound insulators as compared to conventional double wall constructions. The theory indicates that a large improvement in the transmission loss of a double wall can be obtained by using as components single walls with high internal dissipation.