Quality ratings for frequency-shaped peak-clipped speech: results for listeners with hearing loss.

Peak clipping is a common form of distortion in hearing aids and can reduce the subjective quality of the amplified speech. In a previous study involving listeners with normal hearing (Kates & Kozma-Spytek, 1994), the effect of peak clipping on speech quality ratings was studied using sentence test materials that were filtered using three different frequency response contours and then clipped at four different clipping levels. The present study extends the quality ratings to include those from a group of listeners having moderate to profound hearing impairments. The experimental results indicate that the clipping level, and the interaction of the frequency-response shaping with the clipping level, significantly affects speech quality. It is also shown that the distortion effects on speech quality for the listeners with impaired hearing can be modeled by a distortion index computed from the magnitude-squared coherence of the speech-processing system in response to a shaped-noise input signal. The distortion-index weights derived for the group of listeners with impaired hearing, however, differ substantially from those derived for listeners with normal hearing, and substantial inter-listener variation was also observed.

Acoustic-phonetic context considerations for speech recognition testing of hearing-impaired listeners.

OBJECTIVE: To test for differences in the identification of consonants in carrier sentences versus in VCVs extracted from the sentences, as a function of listeners' hearing-loss categories: moderate, severe, profound. To examine whether pauses inserted in the sentences will facilitate identification of the consonants. DESIGN: Voiced stops and fricatives were identified by 11 listeners with moderate hearing losses and by 7 listeners with severe losses (between subjects design) for the conditions of consonants in sentences and in VCVs extracted from the sentences (repeated measures). Nine of these listeners also identified the consonants in the sentences with pauses. Six normal-hearing listeners were tested for the consonants in the extracted VCVs and the sentences. Voiceless stops and fricatives were identified by 4 listeners with profound losses, 18 with severe losses, and 8 with moderate losses (between subjects) for the conditions of extracted VCVs and the sentences (repeated measures). All listeners were selected on the basis of their hearing levels. RESULTS: The listeners with moderate to severe hearing loss identified the voiced stops and fricatives more poorly when the syllables were in the carrier sentences than when extracted. Insertion of the pauses in the sentence did not improve performance significantly. The normal hearing listeners showed no differences in consonant identification between the two conditions, perhaps due to "ceiling effects". The voiceless stops and fricatives were also identified more poorly when in the extracted VCVs than in the carrier sentences by other listeners with moderate to profound hearing loss. CONCLUSIONS: Listeners with moderate or greater hearing loss can show poorer identification of consonants that are embedded midway in carrier sentences than when the acoustically identical consonants are in VCVs extracted from the sentences. The performance reduction for the consonants in sentences is not relieved from insertion of brief artificial pauses in the sentences. Further research is needed determine whether hearing-impaired listeners' identification of consonants in target words of clinical word recognition tests is facilitated when the words are extracted from carrier phrases.

Phonetics of intervocalic consonant perception: retrospect and prospect.

This article reviews the production characteristics and perceptual cues of intervocalic consonants as a background for acoustic studies of consonant perception in fluent speech. Data show that in conversation intervocalic consonants occur much more commonly than consonants in initial or final position; all phonetic features are strongly represented. Production characteristics of intervocalic consonants are seen to depend on the tempo and rhythmic conditions of the syllables of which they are components. At a moderate tempo, consonants in syllable-final position combine with the onset consonant of the following syllable. This affects durational characteristics and may be explained by higher energy efficiency of CV units in production. Phonological phenomena are related to the shifts in syllable position and the temporal compensations of intervocalic consonant production. Studies of consonant perception in fluent context have dealt with tempo of utterance, position in word, and rhythmic pattern, as well as phonemic context. Major phenomena are effects of coarticulation, invariance in consonant perception, and cue interaction and masking. Much evidence suggests a dominance of the perceptual cues in the CV portion of VCCV and VCV sequences. We suggest that exploration of perception variables that affect consonants in fluent context would be expedited by reorienting experimental procedures to employ listener adjustment of stimuli, instead of the traditional phoneme identification and discrimination procedures with large sets of constant stimuli. Most of the relevant literature deals with stop consonants. Lateral, rhotic, and nasal consonants also deserve intensive study because of their very frequent occurrence. Theoretical issues of phoneme perceptual invariance and motor vs. auditory theory of perception are discussed in relation to proposed experiments which vary syllable tempo and stress pattern.

Temporal and spectral cue use for initial plosive voicing perception by hearing-impaired children and normal-hearing children and adults.

The influence of voice-onset time (VOT) and vowel-onset characteristics on the perception of the voicing contrast for initial plosive consonants was examined for hearing-impaired children, and normal-hearing children and adults. Listeners identified spoken 'DAD'--'TAD' stimuli controlled for VOT and vowel onset characteristics. Only six of 16 hearing-impaired children appropriately identified the exemplar DAD and TAD stimuli used as endpoints of VOT continua. For this group of six hearing-impaired children, a longer VOT than for the normal-hearing listeners was required to elicit /t/ rather than /d/ percepts. The VOT region of perceptual cross-over in labelling widened progressively from normal-hearing adults to normal-hearing children to hearing-impaired children. Generally, longer VOTs were required to yield /t/ perception in the context of the DAD vowel than with the TAD vowel. These 'vowel stem' effects on VOT boundary were inconsistent for the hearing-impaired children, and weaker for the normal-hearing children than for the adults. These spoken stimuli produced results for VOT cue use that generally parallel those obtained in studies with synthetic stimuli.

Spectral cues to perception of /d, n, l/ by normal- and impaired-hearing listeners.

The alveolar consonants /d, n, l/ occur frequently in intervocalic position in conversational speech but have received little study for differences in their acoustic cues. Impaired- and normal-hearing listeners were investigated for use of consonant-segment versus transition-segment cues to recognition of /d, n, l/ in /aeCae/ tokens extracted from sentences. To examine the cues' contribution to /d, n, l/ recognition, the segments were degraded singly or in combinations in the tokens as follows: [aeC] or [Cae] transitions were replaced by adjacent pitch periods from the respective vowels; the consonant segments were replaced by silence or by a synthetic consonant approximating the summed low-frequency spectra of the /d, n, l/ murmurs. The results with normal-hearing listeners showed that the presence of any one of the three segments, [aeC] transition, [Cae] transition, or natural consonant segment, supported a moderate to high level of /d, n, l/ recognition, depending on the phoneme. In contrast, the severely hearing-impaired listeners' consonant recognition was poor on the basis of transition information, but better in the presence of the natural consonants. The /aeCae/'s with the synthetic consonant yielded chance level performance for the hearing-impaired listeners but good consonant recognition for the normal-hearing listeners--a further indication that cues in the transitions were quite useful for the normal-hearing group but not for the hearing-impaired group.

VCVs vs CVCs for stop/fricative distinctions by hearing-impaired and normal-hearing listeners.

Moderately to profoundly hearing-impaired (n = 30) and normal-hearing (n = 6) listeners identified [p, k, t, f, theta, s] in [symbol; see text], and [symbol; see text]s tokens extracted from spoken sentences. The [symbol; see text]s were also identified in the sentences. The hearing-impaired group distinguished stop/fricative manner more poorly for [symbol; see text] in sentences than when extracted. Further, the group's performance for extracted [symbol; see text] was poorer than for extracted [symbol; see text] and [symbol; see text]. For the normal-hearing group, consonant identification was similar among the syllable and sentence contexts.

Burst and transition cues to voicing perception for spoken initial stops by impaired- and normal-hearing listeners.

The use of cues to voicing perception of initial stop consonants in multiple spoken syllables was studied for moderately/severely hearing-impaired (n = 43) and normal-hearing listeners (n = 12). The test stimuli were ten utterances each of the syllables/baed, gaed, daed, paed, kaed, taed/. The utterances were analyzed acoustically to confirm the presence of certain cues to initial-stop voicing, namely, differences in voice onset time (VOT), aspiration, and vowel-onset values of the first formant and of fundamental frequency (fo). Test conditions were prepared in which different portions of the syllable onsets were either deleted or interchanged for voicing-cognate syllables. Also the fo contour was flattened for syllable pairs via analysis/synthesis using linear predictor code (LPC) processing. The results confirmed that VOT was a strong voicing cue for both the hearing-impaired and normal-hearing listeners. When the aspirations of the voiceless stops were inserted between the release and the vowel of the voiced-stop syllables, the normal-hearing listeners perceived voiceless stops predominantly. The transition portions of the vowel onsets in burstless /baed, gaed, daed/ contained strong cues for voicing perception of /b,g,d/. The hearing-impaired listeners seemed less sensitive than the normal-hearing listeners to the aspiration-presence and the vowel-onset cues. The fo difference at vowel onset appeared to have no cue value for either group of listeners.

Speech-cue enhancement for the hearing impaired: amplification of burst/murmur cues for improved perception of final stop voicing.

A further study is reported from a program of research exploring the improvement of speech perception by hearing-impaired persons via enhancement of acoustic features of consonants (10,11). Enhancements were applied to certain acoustic segments of consonants, segments known to be useful in consonant perception by normal-hearing persons but often not for persons with severe/profound hearing losses. The consonants were /k/, /t/, /g/, and /d/ located as the final phoneme in /baeC/ words; the voicing feature difference of /k/ versus /g/ and /t/ versus /d/ was the focus of study. The results showed that stop voicing perception improved to at least 90 percent for 3/4 of the listeners when the voiced murmur segments during /d/ and /g/ and the release bursts of /t/ and /k/ were amplified above their natural levels. The audibility of the enhanced segments generally explained differences between the listeners who showed large versus minimal improvements. One training session for stop voicing perception with the cue-enhanced words seemed sufficient to effect maximum performance improvement.

Speech cue enhancement for the hearing impaired: I. Altered vowel durations for perception of final fricative voicing.

Hearing-impaired persons with reduced voicing perception for final fricatives were studied for improvement in perception via training or enhancement of the vowel duration cue. Ten utterances each of the syllables /baef, baes, baev, baez/ composed three stimulus sets that were graded for the magnitude of the vowel duration cue. These sets were used differentially among 25 hearing-impaired listeners to train and test voicing perception. When enhanced, the vowels of the utterances were lengthened before voiced fricatives and shortened before voiceless fricatives. Training for the utterances with unmodified vowels yielded a small performance improvement for a subgroup of 14 listeners. In comparison, this subgroup showed considerably more improvement for the utterances with vowel duration enhancements. Another subgroup of 5 listeners displayed good fricative voicing perception independent of training or the vowel duration enhancements. A remaining subgroup of 6 listeners manifested poor use of the maximum unmodified vowel duration cues, even subsequent to training. This group showed a significant improvement in perception only for the voiced fricatives when extra duration enhancement and further training were used. This subgroup also demonstrated the poorest performance for a measure of vowel duration discrimination. Among the auditory variables examined, the listeners' tone thresholds at 250 Hz showed the highest relation to perception of fricative voicing.

Some rehabilitative considerations for future speech-processing hearing aids.

Improvements in speech-recognition and speech-processing technology promise eventual ways of assisting the hearing impaired by automatically enhancing the audibility of critical speech segments or distinctive features. Some results of applying enhancement techniques are summarized and procedures are proposed for selecting the speech sounds requiring enhancement, the degree of amplification, and the training that hearing-impaired listeners might need.

Perceptual cues to the voiced-voiceless distinction of final fricatives for listeners with impaired or with normal hearing.

Cues to the voicing distinction for final /f,s,v,z/ were assessed for 24 impaired- and 11 normal-hearing listeners. In base-line tests the listeners identified the consonants in recorded /d circumflex C/ syllables. To assess the importance of various cues, tests were conducted of the syllables altered by deletion and/or temporal adjustment of segments containing acoustic patterns related to the voicing distinction for the fricatives. The results showed that decreasing the duration of /circumflex/ preceding /v/ or /z/, and lengthening the /circumflex/ preceding /f/ or /s/, considerably reduced the correctness of voicing perception for the hearing-impaired group, while showing no effect for the normal-hearing group. For the normals, voicing perception deteriorated for /f/ and /s/ when the frications were deleted from the syllables, and for /v/ and /z/ when the vowel offsets were removed from the syllables with duration-adjusted vowels and deleted frications. We conclude that some hearing-impaired listeners rely to a greater extent on vowel duration as a voicing cue than do normal-hearing listeners.

Acoustic cues to final stop voicing for impaired-and normal hearing listeners.

Voicing perception for final stops was studied for impaired- and for normal-hearing listeners when cues in naturally spoken syllables were progressively neutralized. The syllables were ten different utterances of /daep, daek, daet, daeb, daeg, daed/ spoken in random order by a male. The cue modifications consisted progressively of neutralized vowel duration, equalized occlusion duration, burst deletion, murmur deletion, vowel-transition interchange, and transition deletion. The impaired subjects had moderate-to-severe losses and showed at least 70% correct voicing for the unmodified syllables. For the voiced stops, vowel-duration adjustment and murmur deletion each resulted in significant reductions in voicing perception for more than one-third of the impaired listeners; all normals showed good performance following neutralization of these cues. For the voiceless stops, large percentages of both listener groups showed decreased voicing perception due to the burst deletion, though a majority of both groups performed well above change even after the vowel-duration adjustment and the burst deletion. When the vowel off-going transitions were exchanged between cognate syllables in given pairs, the effect on voicing perception exhibited by many impaired- and all normal-hearing listeners implicated the vowel transitions as an important additional source of cues to final-stop voicing perception.

Masking of noise bursts by an adjacent vowel for hearing-impaired listeners.

Some masking effects of synthetic vowels on adjacent noise bursts are reported for hearing-impaired listeners. The 200-ms vowels were similar to [a] and [i]; the 50-ms noise bursts were in one of three frequency bands: .5 to 1.5, 1.5 to 4, or 4 to 6 kHz. With voiceless-stop-like temporal intervals between the stimuli, there was little backward or forward masking of the noise bursts by either vowel. Some forward masking occurred under proximate conditions of vowel and burst in time and frequency.