Stop Bursts Research Articles

Articulatory and acoustic signatures of selected monosyllabic words

At previous meetings of this Society, acoustic signatures for selected words from an intelligibility test have been described. The signatures consist minimally of the formant trajectories of the syllable nucleus of each word. The formant frequency information can be augmented by other acoustic information, such as information on nasal murmurs, stop bursts, silences, and frication intervals. A major goal of this work has been to determine the interspeaker variance of the signatures and to evaluate their usefulness in characterizing speech disorders. This report extends the research to include articulatory data obtained by x-ray microbeam. The articulatory signatures take the form of planar motion paths of radiopaque markers attached to the articulators. Articulatory data were obtained from five talkers. Selected results will be presented to illustrate the articulatory signatures and to show their relationships to acoustic signatures of the same words. The primary goal of this report is to describe intraspeaker and interspeaker variances as they relate to the possibility of developing nonnative standards for articulatory-acoustic signatures. [Work supported in part by NIH.]

Acoustic properties for the distinction of Japanese stop consonants

Of all the sounds of speech. stop consonants present the greatest obstacles to automatic speech recognition. A study has been conducted to find acoustic properties that can be exploited for reliable distinction among the stop consonants of Japanese in CV utterances produced by five male speakers of Japanese. Short‐time spectra were measured at the stop burst and at the onset of the following vowel. It was found that the most consistent property for distinguishing velar stops from other consonants is the presence of a dominant peak in the burst spectrum in the vicinity of the second/third formant of the following back/front vowel. On the other hand, the distinction between alveolar and labial stops can be made on the basis of the second formant transition when the stops are followed by a back vowel. This transition is concave for alveolars and almost level or convex for labials. When these stops are followed by a front vowel, however, the overall slope of the burst spectrum can be more reliably used for their distinction than the second formant transition. A similar analysis was also made on the nasal consonants of Japanese. [Work conducted at Research Laboratory of Electronics, Massachusetts Institute of Technology while the author was on leave from University of Tokyo.]

Speaking clearly for the hard of hearing. II: Acoustic characteristics of clear and conversational speech.

The first paper of this series (Picheny, Durlach, & Braida, 1985) presented evidence that there are substantial intelligibility differences for hearing-impaired listeners between nonsense sentences spoken in a conversational manner and spoken with the effort to produce clear speech. In this paper, we report the results of acoustic analyses performed on the conversational and clear speech. Among these results are the following. First, speaking rate decreases substantially in clear speech. This decrease is achieved both by inserting pauses between words and by lengthening the durations of individual speech sounds. Second, there are differences between the two speaking modes in the numbers and types of phonological phenomena observed. In conversational speech, vowels are modified or reduced, and word-final stop bursts are often not released. In clear speech, vowels are modified to a lesser extent, and stop bursts, as well as essentially all word-final consonants, are released. Third, the RMS intensities for obstruent sounds, particularly stop consonants, is greater in clear speech than in conversational speech. Finally, changes in the long-term spectrum are small. Thus, speaking clearly cannot be regarded as equivalent to the application of high-frequency emphasis.

Automatic target generation for vowels

Research in phonetic recognition by computer is of interest because it may offer better performance than systems using larger‐recognition units such as syllables or words. Among several techniques developed to model speech signals as phoneme sequences (e.g., hidden Markov models and other feature‐based methods) is the phonetic‐target model based on the auditory‐perceptual theory of Miller [ASHA Reports 14 (1984)]. In it: (1) talker differences that mask acoustic phonetic correlates are minimized when incoming speech signals are represented as paths in a three‐dimensional space, where X = log(P3/P2), Y = log(P1/R), and Z = log(P2/P1); and (2) phonetic targets defined in this space seem to be uniquely related to the acoustic patterns of phonelike elements. An automatic procedure generated three phonetic targets for the vowels /ɪ/, /ɛ/, and /ʌ/ from a set of training tokens in stop‐vowel‐stop format recorded from two male and two female talkers. When four new talkers were tested, a vowel recognition accuracy of about 98 % was attained. Additionally, a segmentation algorithm isolated the vocalic segments from the stop bursts with 100% accuracy. [NINCDS grant R01‐NS21994‐01.]

Perception of initial glides by impaired‐ or normal‐hearing listeners

An experiment was conducted to study the use of cues to glide consonant identification by hearing‐impaired (n = 22) and normal‐hearing (n = 6) listeners. The syllables /wæk, jæk, bæk, dæk, gæk, æk/ (six different utterances each) were identified for three conditions: (1) unmodified, (2) initial glide murmurs or stop bursts removed, and (3) syllables cut back to the ninth pitch period from the vowel onset. The normal group, listening at 75 or 95 dB SPL, showed 100% glide phoneme perception for the condition of deleted glide murmurs or stop bursts. For this condition performances varied greatly within the hearing‐imparied group (listening at MCL): Near‐normal glide phoneme perception was seen for less than 20% of the impaired listeners (mean 3FA = 72 dB HL), indicating good use of the vowel onset transitions for glide perception by these listeners. Between the conditions: Unmodified versus glide‐murmur/stop‐bursts deleted, the majority of the hearing‐impaired listeners (mean 3FA = 88 dB HL) manifested degraded manner and/or place and/or phoneme perception. For these listeners reduced perception of the duration and/or frequency extent of F1 and/or F2 vowel onset transitions was suggested by the types of consonant confusions evidenced. About 20% of the impaired listeners (mean 3FA = 92 dB HL) showed no differences in performance among the three conditions, indicating inability to distinguish the presence/absence of vowel onset transitions that characterize initial glides.

Acoustic analyses of closant sounds in infants' comfort‐state utterances

Closants, or consonant‐like sounds, were described acoustically using 16‐kHz spectrograms and LPC or FFT analyses based on waveforms sampled at 20 or 40 kHz. The four major closant types were stops, fricatives and affricatives, vibrants or trills, and glides. Compared to similar sounds in adult speech, the closant sounds of the 3‐, 6‐, and 9‐month‐old infants had significant regions of energy in the higher frequencies, i.e., to and above 14 kHz. An infant's production of closants having the same manner and place often was highly variable in spectral and temporal features, even in the case of reduplicated babbling. For example, the occurrence of stop bursts sometimes was highly variable with respect to voice onset or offset. However, there were instances of reliable reproduction of a basic sound pattern or phonetic property. Acoustic features are described for various places of articulation of stops, fricatives, and trills. The discussion of the data emphasizes (1) the major dimensions of acoustic contrast that are used by infants in the first year of life, (2) implications of the spectral data for auditory and speech motor self‐stimulation by normal‐hearing and hearing‐impaired infants, and (3) the continuity and discontinuity of early closant formation with later phonetic development in the preword and first‐word stages. [Work supported by NIH.]

A Modular 125 PS Resolution Time Interval Digitizer for 10 MHz Stop Burst Rates and 33 MS Range

A high resolution multiple stop time interval digitizer is described. It is capable of resolving stop burst rates of up to 10 MHz with an incremental resolution of 125 ps within a range of 33 ms. The digitizer consists of five CAMAC modules and uses a standard CAMAC crate and controller. All the functions and ranges are completely computer controlled. Any two subsequent stop pulses in a burst can be resolved within 100 ns due to a new dual interpolation technique employed. The accuracy is maintained by a high stability 125 MHz reference clock. Up to 131 stop events can be stored in a 48-bit, 10 MHz derandomizing storage register before the digitizer overflows. The experimental data are also given.

Stop bursts and transitions as functionally equivalent, context-sensitive cues

The experiment reported here was primarily designed to test the functional invariance of stop bursts as perceptual cues. Bursts from three sets of nine CV syllables for each stop-consonant place of articulation were transposed within each set and played for listeners. Although some measure of perceptual invariance was found within each of the labial and apical sets, and within two subsets for the velar bursts, the perceptual weight carried by the bursts attained significance in few of the syllables tested. Further, the perceptual weight of the bursts and formant transitions of the syllables tended to be reciprocally related. These two cues appear to be functionally equivalent, context-dependent cues, each contributing to the rapid spectral changes that follow consonantal release. The results indicate the importance of the role of front cavity resonance in signaling place of articulation.

Influence of encoding and acoustic similarity on the ear advantage and lag effect in dichotic listening.

Experiments with simultaneous and time lag dichotic listening conditions were used to test two hypotheses concerning the right ear advantage and lag effect in dichotic listening. One hypothesis is based on the similarity of acoustic spectra, and the other is based on a categorization of speech sounds as being either encoded or not encoded. Natural vowels and consonant-vowel syllables were used to obtain seven different types of speech stimuli: stop vowel syllables, fricative vowel syllables, stop burst noise, fricative noise, stop vowel transitions, fricative vowel transitions and steady state vowels. The presentation conditions were monaural, simultaneous dichotic, and dichotic with interaural time delays of 15, 30, 60, and 90 msec. With monaural presentations, all stimuli were identifiable above chance levels. For the simultaneous dichotic condition, significant right ear advantages occurred for stop vowel syllables, fricative vowel syllables, stop burst noise, and steady state vowels. For the time lag conditions, stop vowel syllables, stop bursts, and fricative noise produced consistent lag effects, but steady state vowels produced consistent lead effects. In general, the results gave stronger support to the hypothesis of acoustic similarity than to the encoding hypothesis in that stop burst noise produced both a right ear advantage and a lag effect whereas consonant-vowel transitions produced neither a right ear advantage nor a lag effect.

The Effect of Varying Labels on the Identification of Speech and Nonspeech Stimuli Presented Dichotically

When stop bursts labeled as speech sounds were presented dichotically, subjects identified them correctly more often with the right ear than with the left. When labeled as noises of varying pitch, they were identified more often with the left ear than with the right. In the experiment to be presented two types of stimuli, stop bursts (/p,k,t/) and narrow-band noise bursts with frequencies centered around 250 Hz, 1000 Hz, and 4000 Hz were presented in both monotic (control) and dichotic (test) conditions. Forty subjects were divided into four groups. Two of the groups were given stop bursts only, the other two noise bursts only. Of each pair, one group was first instructed to identify the stimuli as speech sounds (/p/, /k/, or /t/), the other as noises of different pitch, low, middle, or high. The tests were then repeated with the instructions reversed. A significant right-ear advantage for stop bursts was found when the first instruction called for speech labeling and a non-significant left-ear advantage when the first instruction called for pitch labeling. No significant left-or right-ear asymmetry was found for noise bursts with either the pitch or the speech labeling condition, regardless or order of presentation.

Tape Cutting Experiments with Stop Consonants

In Danish words with /p,t,k,b,d,g/+/i,a,u/, stop bursts, aspirations, and formant transitions have been removed and exchanged. [Danish /p,t,k/ are strongly aspirated (/t/ affricated), /b,d,g/ are voiceless.] 500 stimuli were presented to 21 listeners, who had to identify the words. Main results were: For the distinction between /p,t,k/ and /b,d,g/, the small difference in stop burst is insignificant, whereas the aspiration is decisive (a simple pause between burst and vowel is not sufficient). The duration of the aspiration is important, but the start of the vowel (noisy start of higher formants) may be sufficient for the identification of /p,t,k/, particularly before /a/, even with VOT values of 20–30 msec. For the identification of place of articulation the aspiration (affrication) of /p,t,k/ is sufficient except for /k/+/a/, but the burst is also sufficient except for t. In /b,d,g/ the transition is sufficient if it is extensive (e.g., /ba/, /da/, /ga/, /bi/, /du/); it is not enough that it points to the locus. The burst is necessary not only in velars before rounded vowels, as previously assumed, but also, for example, in /bu, /di/, and /gi/.

Acoustic Properties of Stop Consonants

The two major cues for stop consonants, the burst of the stop release and the formant transitions in the adjacent vowel, were investigated. Detailed energy density spectra of the bursts were prepared. The transitions were studied by means of sonagrams. Possible criteria for identification were developed and tested. In order to assess the efficacy of the two types of cue, perceptual tests were conducted with isolated segments that contained either stop bursts or vowel transitions alone. Common acoustical properties of bursts and formant transitions are noted; differences as well as similarities are discussed in the light of different varieties of pitch judgments.

On Stop Consonants

Short-time energy vs frequency spectra of stop bursts, and formant transitions in the adjacent vowel have been studied. Sound segments consisting of burst or transition alone were submitted to listeners for identification. Results are discussed with the aim of developing possible identifying criteria utilizing cues in the burst spectra and in the transitions. [This work was supported in part by the Signal Corps, the Office of Scientific Research (Air Research and Development Command), the Office of Naval Research, and the National Science Foundation.]

On Acoustical Cues for Stop Consonants

A study has been conducted of the two major cues for stop consonants: the burst of the stop release and the transition of the formants in the adjacent vowels. Detailed frequency vs intensity spectra of the bursts were prepared, while the transitions were studied by means of Sonagraph records. Possible criteria for identification were developed and tested. In order to assess the efficacy of the two major types of cue, perceptual tests were conducted with isolated segments containing either stop bursts or vowel transitions alone. [This work was supported in part by the Signal Corps, the Office of Scientific Research (Air Research and Development Command), and the Office of Naval Research.]