Voice Spectrum Research Articles

Articulatory characteristics of fluent esophageal speakers for selected english consonants

Linguapalatal contact characteristics and lip and mandible coordination of three fluent alaryngeal esophageal speakers were examined during the production of English consonants /p, t, k, s, b, d, g, z/ in CVC syllables (V = /æ/) embedded in a sentence “_____ is a word.” Pseudopalates with 96 electrodes were fabricated for each subject. Small light reflecting beads were placed on the lips and on cantilevers attached to mandibular and maxillary teeth to determine lip and mandible movements. Information on linguapalatal contact, lip and mandible positions, and voice spectrum were collected every 10 msec with the PAGIS instrumental system. Linguapalatal contact patterns were analyzed with reference to those of normal speakers in terms of place and duration of contact. The contact patterns were related to lip and mandible positions for the criterion CVC syllables. The results will be discussed with respect to known functions of speech physiology. [Supported by NIH grant NS-11852-02 and by NIH grant RR-5349 awarded to the first author.]

Zur Frage der quantitativen Bewertung der Tragf�higkeit der Stimme

The efficiency of the voice is said by Winckel (1953) to be especially determined by the energy of the spectral area around 3000 cps. Therefore, Winckel has repeatedly recommended selective measurement of the 3000 cps-component of the voice spectrum for the objective evaluation of vocal efficiency. As this proposal has not yet been supported by comparative studies, we asked 41 normal subjects and 24 professional singers to phonate several tones (vowel “a”) at different frequencies (c [131 cps]; g [196 cps]; c1 [262 cps] in males and c1 [262 cps]; g1 [392 cps]; c2 [523 cps] in females) and intensities (65, 75, 85 and 95 dB SPL). In all voice samples we compared the formant energy in the spectral area of 2500–3000 cps with the amplitude of a lower frequency area (800–1300 cps), which about corresponds to the area of the second formant, by means of filter analysis. Results show that in all persons tested for this study the relative amplitude of the 3000 cps-component grows with increasing vocal intensity, but diminishes with increasing vocal pitch. At intensity levels of 75 dB and 85 dB SPL singers showed at all three pitch levels (except the tone c2 [523 cps] at 75 dB SPL in females) significant smaller differences between the amplitudes of both the selected formant areas, i.e. relatively stronger partials around 3000 cps, than normal subjects. At the intensity level of 65 dB the differences between vocally trained and untrained groups were less pronounced. This finding supports Winckel's suggestion that the efficiency of a voice can be evaluated quantitatively by the measurement of the relative amplitude of the 3000 cps area, but also reveals that for reliable results such measurement has to be done at certain pitch and intensity levels and under constant test conditions only.

Psychological stress evaluators: EMG correlation with voice tremor.

Temporal measurements of frequency changes in the human voice and tremor in the muscles of the vocal area suggest that these phenomena are correlated. A measure of muscular activity was obtained using a non linear filter on the EMG wave. Frequency changes were detected in the third formant of the voice spectrum. The crosscorrelation results indicate that the voice vibrations are forced oscillations generated by central nervous activity, explaining the detection of stress from voice records.

Correction of Helium Speech Distortions by Real-Time Electronic Processing

The use of helium in breathing gas mixtures for deep diving operations causes marked changes in the speech sounds of the divers. The most prominent effect is an elevation of the formant frequencies which, in conjunction with other phenomena, progressively reduces speech intelligibility as depth increases. Our analyses and investigations have shown that to correct these distortions, the following measures are required: 1) provide a passband extending to at least 10 kHz; 2) equalize high frequency losses in vocal output and in microphone response; 3) reduce voice spectrum envelope frequencies to their expected values for 1.0 atm of air; and 4) preserve the glottal rate of the talker. Recent advances in commercially available electronic signal analyzers have made possible the recording of voice spectra averaged over a wide range of time intervals (50 ms-50 s), over a wide range of frequencies (to 20 kHz) and over a total amplitude range exceeding 100 dB. These capabilities are essential in exploring the total extent of distortions that helium introduces. Analyses of helium speech by means of long term average spectra have revealed phenomena not previously reported. An electronic helium speech processor (unscrambler) incorporating appropriate corrective measures has been developed and evaluated. Its performance culminated in successful retrieval of voice intelligence from divers at 1000-ft pressure depth. Objective word intelligibility test results for corrected speech of one talker were 88 percent at 500-ft and 78 percent at 800-ft depth.

Spectrum Conservation and Characteristics of Single-Sideband Phase Modulation

Phase modulation with an analytic signal, which is a Gaussian random process, is examined in order to determine the amount of spectrum conservation that may be achieved by using single-sideband phase modulation (SSB-PM) rather than conventional phase modulation (PM). The autocorrelation function is derived and found to be an analytic signal in terms of the autocorrelation function of the actual modulating signal and its Hilbert transform. When the modulating signal strength is very low, the sideband spectral distribution is the same as that of the actual modulating signal or single-sideband amplitude modulation. As the modulating signal mean-square value is increased, the sideband spectrum broadens and approaches a Gaussian shape. The average power output of an SSB-PM system increases exponentially with input modulating signal strength, while the carrier power remains constant. For the same modulating signal mean-square value, a greater fraction of power is in the one sideband of an SSB-PM system than in the two sidebands of conventional PM. Single-sideband phase or frequency modulation always effectuates spectrum conservation in the continuum when it is compared with conventional phase or frequency modulation on the basis of equal relative sideband power. A Fourier transform computer program is used to generate SSB-PM spectral distributions with varying modulating signal mean-square values, when the modulating signal spectrum is a low-pass rectangular spectrum, a narrowband pass spectrum, and the shape of an average voice spectrum. These examples illustrate the power series formulation of the output spectrum as well as the theoretical analysis of bandwidth.

Mewing cry in a child with the partial deletion of the short arm of chromosome no. 4.

A case is presented of partial deletion of the short arm of the chromosome No. 4, but with a mewing cry, typical of the 5p — deletion syndrome. The clinical examination revealed similar features to those described in other cases of 4p — deletion, namely low birth weight, hypertelorism, facial asymmetry, failure to thrive, mental retardation, beak-shaped nose, low set ears, broad nasal bridge, skeletal anomalies and hypotony. The mewing character of the voice was confirmed by analysis of the voice spectrum. The deleted chromosome was identified by the measurement technique.

Problems of diver communication

The major problems of underwater communication are microphone and earphone construction speech distortion caused by high pressure and/or mixed gas breathing environments, and face mask design. Microphone and earphone elements have recently been developed that are relatively unaffected by increased ambient pressure. These elements are constructed of piezoelectric ceramics encased in a unique rubber housing to prevent water leakage. Preliminary tests indicate that the frequency response of these elements is changed less than 3 dB by pressures exceeding 500 lbf/in <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> gauge. Extensive research in speech distortion under extreme pressures with various gas mixtures includes voice spectrum analysis, study of adaptation changes in speech, and intelligibility measurements of helium speech before and after processing or "unscrambling." These basic data provide information for the design of microphones, masks, and electronic and mechanical means of normalizing distorted speech. Face mask design must include consideration of breathing requirements, as well as speaking requirements. The design of a speaking cavity has, of necessity, been based on empirical data. These designs often conflict with the primary requisite of proper breathing design. Electronic compensation is often employed to correct or improve deficiencies in mask characteristics.

Tone-selective waystation dialing system interconnecting "message" and regular telephone line

Although railway dispatchers' telephone circuits should normally be entirely separate from other telephone facilities, there are times when interconnection is desirable. This paper describes dial equipment which would complete such connections without the aid of an operator, and without using the existing "Type 60" selectors. This system utilizes resonant-reed oscillators and relays in the voice spectrum. It requires little power (no standby power required at the way-station) and has low bridging losses. Tuned "guard" circuits preclude switching response to voice currents.

Improved Multiplex Voice Frequency Carrier Systems Using New Audio Filters

Using resonators built with high Q ferrite-cored inductors and suitably selected plastic film capacitors, improved band-pass and band-elimination filters have been designed which allow the accommodation of duplex high-speed telegraphy channels in a mid-band-speech carrier circuit, with small degradation in voice quality and with negligible cross-talk. It is shown that in such a circuit, the removal of a 400-cps frequency band from the voice spectrum, from about 1100 to 1500 cps, offers the best compromise as far as intelligibility of speech is concerned. The basic elements of a complete system are described. The improved tone channel filters accept, with small distortion and ringing, frequency shifts of ±42.5 cps and allow: 1) the reduction of the usual 170-cps channel spacing to 145 cps, without increase in cross-talk, and 2) the location of 145-cps spaced tones up to 6000 cps in suitable links. For systems with frequency shifts of ±30 cps, similar units with a narrower acceptance band reduce the required channel spacing to only 85 cps.

Statistical Studies of Digital Voice Data

The results of some statistical data processing techniques which have been applied to the study of digital voice data are discussed. Speech signals were analyzed in an 18-channel spectrum analyzer and converted to a digit stream at a data rate of 3000 bits per second. The digital data were recorded on tape and processed on general purpose digital computers by various methods in order to examine the underlying structure of the speech information. Exploratory studies have included the compilation of tables of relative frequencies of occurrence of various voice spectrum pattern modes and groupings. The cross correlatior of voice spectrum amplitude information as a function of both frequency and time was investigated, along with the effect of transforming from frequency spectrum coordinates to a new set of uncorrelated coordinates.

Intelligibility at High Voice Levels and the Use of a Megaphone

A small megaphone was tested for use in direct voice communication high levels of vocal output against backgrounds of noise. Intelligibility scores were determined for 12 talkers, speaking with and without the megaphone, over a range of field voice levels of 70 to 100 db SPL (at 1 m). Nearly the entire over-all acoustical gain produced by the megaphone was realized in improved intelligibility above backgrounds of white noise and low-frequency noise. The acoustical gain of the megaphone permitted a lower voice level, and, hence, delayed the deterioration in intelligibility associated with distortions due to shouting. Frequency shifts in the voice spectrum were measured, with and without the megaphone, over a range of vocal efforts from conversational to maximum shout. Control tests demonstrated that the voice-frequency shifts associated with shouting are not the primary factor in the deterioration of intelligibility; the frequency shifts appear to be only symptomatic of serious vocal distortions in shouting.

Shouting over Noise with a Megaphone

A small megaphone was tested for use in direct voice communication at high voice levels against backgrounds of noise. The voice transmission characteristics of the megaphone were measured. Intelligibility scores were determined for ten talkers, speaking with and without the megaphone, over a range of field voice levels of 70 to 100 db SPL (at 1 meter). A white noise and a low-frequency noise background were employed. Nearly the entire over-all acoustical gain produced by the megaphone was realized in improved intelligibility above both noise spectra. The acoustical gain of the megaphone permitted a lower voice level, and, hence, delayed the deterioration in intelligibility associated with distortions due to shouting. Frequency shifts in the voice spectrum were measured over a range of vocal efforts from conversational maximum shout. Control tests demonstrate that the voice-frequency shifts associated with shouting are not the primary factor in the deterioration of intelligibility; the frequency shifts appear to be only symptomatic of serious vocal distortions in shouting.

Normalization of the Voice Spectrum

Measurement of the spectrum pattern of voice signals by a method that normalizes (i.e., makes the measurement unresponsive) with respect to amplitude fluctuations offers some powerful advantages. The relative amplitude at all frequency positions in the voice spectrum pattern is significantly correlated with the particular speech event; therefore a measurement of the total spectrum pattern offers a means of maximum utilization of the speech signal information. Normalization has the effect of slowing down the rate of change of the signal parameters that specify the voice spectrum, abstracting the common-mode fluctuation due to voice amplitude and specifying it as a separate parameter. Further, a specification of the normalized spectrum patterns of speech offers in principle a more economical specification than one in which formant frequencies are specified as independent parameters. The normalizing method has been reported previously [C. P. Smith, J. Acoust. Soc. Am. 28, 159 (1956)]; some speech data taken with normalization are presented.

Ear-Insert Microphone

It has been observed by several different technical groups that an acoustic transducer placed at the ear can be used as a microphone to pick up the voice of the wearer. This paper presents the results of an investigation to determine the various factors which make possible the use of a microphone in the ear. The investigation, using a laboratory model dynamic ear-insert microphone and a protective headset, determined the relative importance of the two principal paths that conduct sound pressure from the mouth to the outer ear canal: (1) The air path around the head, and (2) the path through the head. The voice spectrum in the outer ear canal, contributed by the path through the head, was also obtained. It appears that the ear-insert microphone could compete with a conventional microphone in front of the lips only in an application where it is difficult to use the conventional microphone because of its size or its interference with other equipment. The investigation forms a basis for further study to determine whether the ear insert microphone has a future as a practical communications device.

Objective Measurement of the Effectiveness of Telephone Sets (La Mesure Objective de l'Efficacité des Appareils Téléphoniques)

This paper describes the physical principles and technical equipment developed by the Centre National d'Etudes des Télécommunications de Paris for evaluating telephone sets objectively. This objective evaluation procedure replaces the subjective evaluations hitherto in normal use. The system consists of an artificial voice that is mechanical or electronic depending on the manner of use (tests in factories or measurements at the home of the subscriber), an artificial mouth, and an artificial ear (a complete and simplified model). The characteristics of this equipment duplicate those of the corresponding human organs as determined for a large number of subjects. The artificial mouth is set in an artificial head. The average spectrum of the artificial voice reproduces that of the natural voice. The artificial ear contains two elements: One is geometrical and mechano-acoustical, and simulates the acoustical impedance of the ear in such a manner that the sound pressure produced by a receiver applied against its cavity is the same as the sound pressure produced in the case of a similar application against an average external auditory meatus. The second, purely electrical, element gives to the detector system of the artificial ear a transfer function that is analogous to that of a natural ear. Experience has shown a very satisfactory correspondence between results of measurements performed with the described equipment and those obtained by a team of well-trained human subjects. Description d'ensemble des principes physiques et des appareils techniques expérimentés par le Centre National d'Etudes des Télécommunications, pour le contrôle objectif des appareils téléphoniques et sa substitution au contrôle subjectif jusqu'ici en service normal. Le système utilisé comporte une voix artificielle, mécanique ou électronique (suivant le mode d'emploi: vérification en usine, ou mesure au domicile de l'abonné), une bouche artificielle et une oreille artificielle (modèle complet et modèle simplifié). Les caractéristiques de ces appareils reproduisent celles des organes correspondant qui ont été déterminées sur un grand nombre de sujets. La bouche artificielle est incorporée à une tête artificielle. Le spectre moyen de la voix artificielle reproduit celui de voix naturelle. En ce qui concerne l'oreille artificielle, elle comporte deux éléments, l'un géométrique et mécano-acoustique qui simule, indirectement, l'impédance acoustique de l'oreille, de telle façon que la pression acoustique, produite par un récepteur appliqué contre son pavillon, soit la même que la pression produite dans le cas d'une application semblable contre le pavillon d'une oreille naturelle moyenne. Le second élément, purement électrique de l'oreille, donne au système détecteur de l'oreille artificielle une loi de fonctionnement analogue à celle de l'oreille naturelle. L'expérience a montré une correspondance très satisfaisante entre les résultats des mesures exécutées avec le dispositif décrit et avec une équipe d'opérateurs bien entraînés.

A Speech Sound Coder

An analyzer has been devised for the direct coding of speech sounds in terms of their characteristic spectral patterns. A conventional set of analyzing filters, with rectifiers and low-pass filters, transforms the voice signal to a set of dc voltages that vary at relatively slow rates. The analyzer transforms these voltages to a group of binary signals that specify the ratio of each filter voltage to the mean or average filter voltage from the set in a continuous measurement, thus providing normalization of the voice spectrum patterns with respect to the amplitude of the voice signal. The analyzer presents the measurement in a matrix having the dimensions of frequency and relative amplitude in decibels; the “occupied” matrix cells represent the spectral pattern of the voice signal at any moment. The transformed patterns are on a logarithmic scale of relative amplitude and are invariant with amplitude changes of the input voice signal over a wide dynamic range. Signals from the matrix “cells” can be summed in various pattern configurations to generate voltages that will serve to automatically detect the occurrence of particular spectral patterns, thus detecting speech sounds in terms of their spectra. A “many-to-one” relationship in the pattern output connections will identify a particular sound whether spoken by an adult or child, whether stressed or unstressed, etc. A thousand such patterns may provide “high-fidelity” voice coding, yet if sampled thirty times a second would require only 300 bits a second to specify the pattern sequences by a digital code. In addition to pattern matching uses, the matrix can drive an indicator matrix of neon bulbs, to generate a running visual display of the normalized spectrum of the voice signal, on a logarithmic scale of relative amplitude. The analyzer is not yet completed, but some of the novel circuit features will be discussed, and possible analogies of the system with the hearing mechanism will be conjectured.