The effects of increasing consonant/vowel intensity ratio on speech loudness.

It was hypothesized that speech loudness may be primarily determined by the level of the vowel and that, as a consequence, high positive consonant/vowel intensity ratios (C/V ratios) could be tolerated by hearing-impaired listeners with possible improvement in intelligibility. The present study was concerned with the effects of high C/V ratios on the loudness of speech as a necessary first step prior to more detailed studies of loudness tolerance and intelligibility. Recordings of four CVC monosyllables were digitized and one of the consonants in each word was selected for amplification relative to a constant vowel level. For each word a set of seven tokens was prepared representing a range of C/V ratios from approximately -20 dB to 9 dB. The loudness of each token was obtained through a loudness matching task involving a standard word presented at 90 dB SPL. In addition, sets of nonspeech stimuli were created to approximate the C/V ratios represented in two of the monosyllables. Loudness of nonspeech tokens was obtained using the same loudness matching paradigm. It was found that high C/V ratios had no appreciable effect on speech loudness. (The nonspeech stimuli gave similar results, however, so it was not possible to conclude that speech was unique in that respect.) The findings in general are encouraging for the further study of the influence of C/V ratio on intelligibility and the eventual incorporation of C/V processing into digital hearing aids.

The intelligibility of time-domain-edited esophageal speech.

The intelligibility of esophageal speech has been shown to be significantly lower than that of normal laryngeal speech. The current study investigated the possibility of enhancing the intelligibility of esophageal speech by manipulating samples in the time domain. Specifically, injection noises and nonphrasal pauses were digitally edited from the speech samples of 5 esophageal talkers. Twenty-five sentences were selected and edited in the time domain and presented to 15 naive listeners who were instructed to write down the words that they heard. The percentage of correct words heard for each sentence was determined and compared across listeners, sentences, and talkers. The overall effect of the editing was a small but significant gain in the intelligibility of the esophageal speech. The improvement in intelligibility, however, depended on the individual talker, the speech material, and the number of editing changes made to a particular sample.

An evaluation of residue features as correlates of voice disorders.

Two experiments were conducted to assess the correlations of residue features with some perceptual properties of voice disorders. First, 90 samples of the vowel /a/ produced by patients with various vocal pathologies were analyzed to obtain the residue features, and severity judgments of these vowel samples were obtained. The results of linear multiple regression analysis indicated that the features were highly correlated with the severity ratings. Second, an attempt was made to correlate the residue features with voice qualities. The features were calculated for the vowel /a/ produced by patients with vocal nodules, vocal fold paralysis, and vocal polyps and by normal talkers. Each vowel sample was rated on ten scales of voice quality. The results revealed high correlations among the quality scales so that discrete subject groups could not be formed. Thus, residue features may be useful in assessing the degree of vocal impairment, but their use as correlates of voice quality must await further research.

Intraoral air pressure as a feedback cue in consonant production.

The effects of oral-sensory deprivation on the production of consonants was studied using narrow phonetic transcriptions and measurements of intraoral air pressure and duration. The speech materials were 20 bisyllabic words produced both in isolation and in sentences, and sentences that included words with 34 stop variants. These utterances were produced by four normal talkers and by the same talkers when deprived of oral sensation. The state of oral-sensory deprivation was induced by a series of mandibular, infraorbital, and palatal injections of 2 percent xylocaine. The talkers also scaled levels of effort used to produce the syllables /pa/ and /ba/, both with and without the anesthetic. In general, after the administration of the anesthetic, the characteristic tongue carriage of the talkers was shifted posteriorly, the rate of speech was slower, and there were minor imprecisions in articulation consisting primarily of alterations in lip and tongue activity. In addition, consonants were produced with slightly greater intraoral air pressures and longer durations. The talkers had no difficulty in scaling levels of effort in either the normal or the anesthetic conditions, and maintained a linear relationship between effort and intraoral air pressure in both conditions. The results suggest that the talkers used more effort in producing speech in the anesthetic condition and are untenable with the idea that intraoral air pressure constitutes an important feedback parameter in controlling articulation. The speech produced by the talkers while anesthetized (that is, while without sensation in the mouth) was reasonably precise and must be postulated to have been under the control of a pressure-sensing system other than a closed feedback loop.

Temporal characteristics related to the discrimination of stutterers' and nonstutterers' speech samples.

Measurements of the difference in average speaking rate, average number of pauses, average pause duration, and average duration of the vowels that received primary stress were obtained from the speech samples of 35 stutterers and 35 nonstutterers. The samples had been screened to ensure that they contained no instances of overt stuttering, audible respirations, or inappropriate voicing. The measurements were used as the predictors in multiple linear regression analyses. The criterion variable was the average percent-correct discriminations of 40 subjects who listened to the samples in pairs and indicated which member of each pair was the stutterer. The results showed that the difference in speaking rate combined with either pause measure accounted for approximately 70% of the variance in the listeners' responses. The findings indicate that speaking rate and pauses are potential perceptual cues for listeners attempting to discriminate the speech of stutterers from that of nonstutterers.

Effects of segment and pause manipulations on the identification of treated stutterers.

This study examined the hypothesis that reading rate affects the identification of treated stutterers. Thirty-two pairs of recorded speech samples, in which one member of the pair was a treated stutterer and the other was a nonstutterer, were available from previous research. Listeners had been able to distinguish readily between the members of these pairs. For each pair of samples, the durations of the treated stutterer's segments and pauses were adjusted to match those of the nonstutterer as closely as possible by means of a computer-based waveform editor. A test tape was prepared that included the 32 original pairs of stimuli, the 32 edited pairs, and 64 pairs of foils. Listeners were required to indicate which member of each pair was the treated stutterer. Analysis of the responses indicated that the listeners' ability to distinguish between talkers was significantly reduced for the edited stimulus pairs. The results imply that the rate used by treated stutterers must be critically evaluated if the goal of therapy is the production of perceptually "normal" speech.

Effects of consonantal context on vowel lipreading.

The effects of consonantal context on vowel lipreading were assessed for 30 adults with mild-to-moderate sensorineural hearing loss who lipread videotape recordings of two female talkers. The stimuli were the vowels /i,I,alpha,U,u/ in symmetric CVC form with the consonants /p,b,f,v,t,d,f,g/ and in the asymmetric consonantal contexts /h/-V-/g/, /w/-V-/g/, /r/-V-/g/. Analyses of the confusion matrices from each talker indicated that vowel intelligibility was significantly poorer in most contexts involving highly visible consonants, although the utterances of one talker were highly intelligible in the bilabial context. Among the visible contexts, the fricative and labiodental contexts in particular produced the lowest vowel intelligibility regardless of talker. Lax vowels were consistently more difficult to perceive than tense vowels. Implications for talker selection and refinement of the concept of viseme were drawn.

Constancy of relative timing for stutterers and nonstutterers.

Fifteen stutterers and 15 nonstutterers read a 120-word passage five times in succession. From the stutterers' readings, sentences were selected for analysis that were produced fluently in the first and the fifth reading. The sentences surrounding the target utterance in the first reading, however, contained instances of stuttering although the surrounding sentences in the fifth reading were fluent. The same utterances were selected from the first and fifth readings produced by the nonstutterers, but the surrounding sentences were fluent for both samples. Four separate relative timing ratios were defined by measuring an acoustic period and an acoustic latency and dividing the period by the latency. Analysis of the ratios revealed no significant differences between the groups in spite of the rate changes that occurred between the readings. The data indicate that not all aspects of a stutterer's speech are affected by the stuttering, and that relative timing may be a critical parameter for the production of fluent utterances.

Perception of synthetic visual consonant-vowel articulations.

Synthetic speech-like articulations were presented to adult subjects via the visual modality, following the classic categorical perception experimental paradigm (Liberman, Harris, Hoffman, & Griffith, 1957). Animations were generated on a computer-based graphics system. Stimuli consisted of representations of the syllables /ba/, /va/, and /wa/; as well as 6 linearly interpolated intermediate stimuli between each of the possible exemplar pairs, resulting in three 8-item continua. Three sets of observations were obtained for these stimuli. First, for each continuum, labeling data were obtained in which the subject assigned one or the other exemplar label to each of the stimuli. Next, ABX discrimination data were obtained for each continuum. In the final task, subjects assigned a rating of one through nine to each animation indicating the extent to which it was like the exemplar syllables. Although the labeling functions showed rather abrupt transitions from one response category to the other, the peaks in the discrimination functions did not coincide with the category boundaries. Further, the mean rating functions were relatively linear, and the distribution of rating responses revealed unimodal distributions whose peak locations differed depending on the stimulus.

Speech naturalness ratings of treated stutterers.

This investigation compared the speech naturalness ratings of perceptually fluent speech samples produced by nonstutterers and stutterers who had been treated in six different therapy programs. The treated stutterers were then divided into two groups based on ease of identification using stutterer/nonstutterers judgments. The fluent speech of one group of stutterers was easily recognized as being produced by stutterers. The second group produced fluent speech that was difficult to distinguish from that used by normal talkers. Finally, pretreatment severity ratings (i.e., mild, moderate, and severe) were compared with posttreatment naturalness scores. Results indicated that a statistically significant difference existed between the naturalness ratings of the nonstutterers and the treated stutterers. Also, a significant difference was found between the naturalness ratings of the easy-to-identify stutterers. Finally, no difference was found in the posttreatment naturalness ratings of stutterers rated as mild, moderate, and severe before treatment.

Phoneme boundaries in normal and hearing-impaired ears.

Two experiments are reported which explore variables that may complicate the interpretation of phoneme boundary data from hearing-impaired listeners. Fourteen synthetic consonant-vowel syllables comprising a/ba-da-ga/ continuum were used as stimuli. The first experiment examined the influence of presentation level and ear of presentation in normal-hearing subjects. Only small differences in the phoneme boundaries and labeling functions were observed between ears and across presentation levels. Thus monaural presentation and relatively high signal level do not appear to be complicating factors in research with hearing-impaired listeners, at least for these stimuli. The second experiment described a test procedure for obtaining phoneme boundaries in some hearing-impaired listeners that controlled for between-subject sources of variation unrelated to hearing impairment and delineated the effects of spectral shaping imposed by the hearing impairment on the labeling functions. Labeling data were obtained from unilaterally hearing-impaired listeners under three test conditions: in the normal ear without any signal distortion; in the normal ear listening through a spectrum shaper that was set to match the subject's suprathreshold audiometric configuration; and in the impaired ear. The reduction in the audibility of the distinctive acoustic/phonetic cues seemed to explain all or part of the effects of the hearing impairment on the labeling functions of some subjects. For many other subjects, however, other forms of distortion in addition to reduced audibility seemed to affect their labeling behavior.

Binaural loudness summation in the hearing impaired.

Binaural loudness summation was measured using three different paradigms with 10 normally hearing and 20 bilaterally symmetrical high-frequency sensorineural hearing loss subjects. An adaptive paradigm and a loudness matching procedure measured summation at the lower and upper level of comfortable loudness and the loudness discomfort level (LDL). Monaural and binaural LDLs also were obtained with a clinical procedure designed to select maximum output of hearing aids. Stimuli for all three tasks consisted of 500- and 4000-Hz pure tones and a speech spectrum noise. Binaural summation increased with presentation level using the loudness matching procedure, with values in the 6-10 dB range. Summation decreased with level using the adaptive paradigm, and no summation was present with the clinical LDL task. The hearing-impaired subjects demonstrated binaural summation that was not significantly different from the normally hearing subjects. The results suggest that a bilaterally symmetrical sensorineural hearing loss does not affect binaural loudness summation. The monaural and binaural dynamic range widths were similar, and the LDL results suggest that binaural loudness summation may not be an important factor in selecting maximum output of hearing aids.

Description and validation of an LDL procedure designed to select SSPL90.

A new procedure is described for measuring loudness discomfort levels (LDLs) for the purpose of selecting SSPL90 characteristics of hearing aids. The person is seated in sound field wearing a high-output hearing aid (with a known amount of 2 cm3 coupler gain) connected to a personal earmold. The loudness of frequency-specific signals is rated from a series of loudness category descriptors. The LDL is defined in terms of SPL developed in a 2 cm3 coupler, thus making selection of SSPL90 from hearing aid specification sheets practical. Experiments on LDL stability over time and validation of the SSPL90 selection are reported.

Training auditory-visual speech reception in adults with moderate sensorineural hearing loss.

A new method of training auditory-visual speech reception is described and evaluated on an experimental group of 12 hearing-impaired adult patients. The method involves simultaneous, live presentation of the visible and acoustic components of the therapist's speech, where the acoustic signal is degraded under the therapist's control with a voice-activated switch. Pre- and post-training performance was assessed with an auditory-visual sentence recognition task. The performance of the experimental group, who received 10 hours of individual training, is described and compared to a control group who received a traditional aural rehabilitation program and to a group of normals who received no training. The experimental training resulted in significantly greater improvement than the control group. A description of the training, including rationale and suggestions for implementation in a clinical setting, is provided.

Consonant similarity judgments by normal and hearing-impaired listeners.

This investigation attempted to identify listener strategies or perceptual modes that might be adopted by hearing-impaired listeners when making similarity judgments among pairs of speech sounds. Further, an attempt was made to describe the relationship between similarity judgments and auditory confusions for such listeners. Subjects provided similarity ratings and recognition responses to consonant pairs. The resulting similarity judgments were organized into a variety of similarity matrices and analyzed via multidimensional scaling and hierarchical clustering, as well as by traditional descriptive and interpretative statistics. The analyses of the similarity ratings between consonants showed that hearing-impaired listeners apply phonemic labels to the stimuli and base their ratings on these labels rather than on the unlabeled acoustic characteristics of the speech sounds. Analysis of the recognition data indicated that those consonants which are most confused are not necessarily the most conceptually similar to the listener.

Formant frequencies of stuttered and fluent vowels.

The formant frequencies of 15 adult stutterers' fluent and disfluent vowels and the formant frequencies of stutterers' and nonstutterers' fluent vowels were compared in an F1-F2 vowel space and in a normalized F1-F2 vowel space. The results indicated that differences in formant frequencies observed between the stutterers' and nonstutterers' vowels can be accounted for by differences among the vocal tract dimensions of the talkers. In addition, no differences were found between the formant frequencies of the fluent and disfluent vowels produced by the stutterers. The overall pattern of these results indicates that, contrary to recent reports (Klich & May, 1982), stutterers do not exhibit significantly greater vowel centralization than nonstutterers.

Examination of two issues concerning functional gain measurements.

The test-retest variability of aided sound-field thresholds and the feasibility of using corrected 2-cm3 coupler gain values to predict individual functional gain were examined. Test-retest data were used to generate critical differences (in dB) for statistical significance between two aided sound-field thresholds. To be significantly different at the .05 level, two aided thresholds would have to differ by greater than 15 dB. Functional gain and 2-cm3 coupler gain were compared for 20 subjects using careful measurement procedures. Individual variability in the difference between functional gain and 2-cm3 coupler gain was reduced substantially from previous studies and indicates that corrected coupler gain measurements may have some clinical utility.

EMG biofeedback in the treatment of hyperfunctional voice disorders.

In this study, six patients who were judged to produce speech with excessive laryngeal tension participated in 14 30-min biofeedback training sessions. The EMG signal, measured by bipolar surface electrodes placed over the cricothyroid region, was amplified, integrated, and used to control the output of a voltage controlled oscillator (VCO) and a noise generator. A voltage comparator was used as a threshold device. Thus, the noise generator was activated when the EMG activity exceeded the value selected on the comparator, and the VCO was activated when the EMG signal was below threshold. The results demonstrated that three of the subjects reduced the laryngeal EMG activity used during speech production with a concomitant improvement in voice quality. The voice quality of the three remaining subjects was unchanged. Suggestions are made for modifying the equipment and procedures used in this study for routine clinical use of EMG biofeedback.

A comparison of the effects of hearing impairment and acoustic filtering on consonant recognition.

The relationship between consonant recognition under conditions of acoustic filtering and hearing loss was studied in subjects with unilateral bearing impairments. Using a procedure involving suprathreshold loudness balance between ears, a spectrum shaper was used to match the (suprathreshold) audiometric configuration of the impaired ear. Consonant recognition data were obtained from the impaired ear and from the normal ear listening through the spectrum shaper. To the extent that consonant recognition was similar in the two ears, the effect of the patient's hearing impairment on phoneme identification could be related to the audiometric configuration. A comparison between consonant recognition scores for the impaired ears and for the ears listening through the spectrum shaper revealed large individual differences among subjects. Mean consonant recognition ability, however, was generally lower in the impaired ear. Although overall consonant recognition and the error probabilities for individual consonants tended to be different between ears, the patterns of feature recognition were quite similar.

Visual biasing of normal and impaired auditory speech perception.

Intersensory biasing occurs when cues in one sensory modality influence the perception of discrepant cues in another modality. Visual biasing of auditory stop consonant perception was examined in two related experiments in an attempt to clarify the role of hearing impairment on susceptibility to visual biasing of auditory speech perception. Fourteen computer-generated acoustic approximations of consonant-vowel syllables forming a /ba-da-ga/ continuum were presented for labeling as one of the three exemplars, via audition alone and in synchrony with natural visual articulations of /ba/ and of /ga/. Labeling functions were generated for each test condition showing the percentage of /ba/, /da/, and /ga/ responses to each of the 14 synthetic syllables. The subjects of the first experiment were 15 normal-hearing and 15 hearing-impaired observers. The hearing-impaired subjects demonstrated a greater susceptibility to biasing from visual cues than did the normal-hearing subjects. In the second experiment, the auditory stimuli were presented in a low-level background noise to 15 normal-hearing observers. A comparison of their labeling responses with those from the first experiment suggested that hearing-impaired persons may develop a propensity to rely on visual cues as a result of long-term hearing impairment. The results are discussed in terms of theories of intersensory bias.