Acoustic dimensions of hearing-impaired speakers' intelligibility: segmental and suprasegmental characteristics.

Regression and principal components analyses were employed to study the relationship between 28 segmental and suprasegmental acoustic parameters of speech production and measures of speech intelligibility for 40 severely to profoundly hearing-impaired persons in an effort to extend the findings of Metz, Samar, Schiavetti, Sitler, and Whitehead (1985). The principal components analysis derived six factors that accounted for 59% of the variance in the original 28 parameters. Consistent with the findings of Metz et al., a subsequent regression analysis using these six factors as predictor variables revealed two factors with strong predictive relationships to speech intelligibility. One factor primarily reflected segmental production processes related to the temporal and spatial differentiation of phonemes, whereas the other primarily reflected suprasegmental production processes associated with contrastive stress. However, the predictive capability of the present factor structure was somewhat reduced relative to the findings of Metz et al. (1985). Data presented indicate that the populations sampled in the two studies may have differed on one or more dimensions of subject characteristics. Considered collectively, the present findings and the findings of Metz et al. support the tractability of employing selected acoustic variables for the estimation of speech intelligibility.

Articulatory dimensions of hearing-impaired speakers' intelligibility: evidence from a time-related aerodynamic, acoustic, and electroglottographic study.

Regression and principal components analyses were employed to study the relationship between 30 aerodynamic speech parameters and the speech intelligibility of 40 severely to profoundly hearing-impaired speakers. Regression analysis on the original 30 aerodynamic variables revealed that speech intelligibility was predicted by a cognate-pair voice onset-time difference measure and a measure of the stability of the volume-velocity rise time. Principal components analysis of the 30 independent variables derived seven factors that accounted for 84.3% of the variance in the original 30 parameters. Subsequent regression analysis using the seven factors as predictor variables revealed four factors with independent relationships to speech intelligibility. These included a factor that reflected cognate-pair voice onset-time distinctions, a factor that reflected cognate-pair peak volume-velocity distinctions, and two other factors, which reflected production stability of temporal distinctions between cognate pair members.

Acoustic dimensions of hearing-impaired speakers' intelligibility.

Regression and principal components analyses were employed to study the relationship between three measures of speech intelligibility and 12 segmental, prosodic, and hearing ability parameters in 20 severely to profoundly hearing-impaired speakers. Regression analyses on the original 12 parameters revealed that cognate pair voice onset time differences and mean sentence duration strongly predicted speech intelligibility based on readings of isolated word and contextual speech material. A principal components analysis derived four factors that accounted for the majority of the variance in the original 12 parameters. Subsequent regression analyses using the four factors as predictor variables revealed two factors with strong relationships to the speech intelligibility measures. One factor primarily reflected segmental production processes related to the temporal and spatial differentiation of phonemes, whereas the other factor reflected prosodic features and production stability. These results are consistent with prior research that suggests independent primary and secondary roles for segmental and prosodic speech characteristics, respectively, in determining intelligibility in severely to profoundly hearing-impaired speakers.

Prediction of contextual speech intelligibility from isolated word intelligibility measures.

The purpose of this study was to evaluate the efficacy of two nonlinear formulae posited by Boothroyd (1978) for prediction of contextual intelligibility from measures of isolated word intelligibility with four different sets of speech intelligibility data. Both of Boothroyd's formulae, Pc = 1 - (1 - Pi)2 and Pc = 1 - (1 - Pi)1.5 (1 + Pi), provided excellent fit of measured contextual intelligibility with predicted contextual intelligibility for all sets of data. Attempts to improve correlation coefficients between measured and predicted contextual intelligibility by iterating Boothroyd's suggested exponents did not significantly improve the predictive ability of the original formulae.

Direct magnitude estimation and interval scaling of stuttering severity.

The appropriateness of direct magnitude estimation and interval scaling for assessing stuttering severity was investigated by determining whether the continuum of the stutterers' judged severity was prothetic or metathetic. As operationally defined by Stevens, prothetic continua show a curvilinear relation between magnitude estimates and interval scale values of the same set of stimuli, whereas metathetic continua show a linear relation between these scale values. The stuttering severity of 20 stutterers was scaled by three groups of 15 listeners who used interval scaling, direct magnitude estimation with standard/modulus, and direct magnitude estimation without standard/modulus. The results indicated that the two sets of direct magnitude estimation scale values were related to the interval scale values in the curvilinear fashion that is typical of prothetic continua. These findings suggest that direct magnitude estimation is preferable to interval scaling for measuring stuttering severity.

Contextual effects in the measurement of hearing-impaired speakers' intelligibility.

Previous research has indicated that hearing-impaired speakers' intelligibility scores are better when sentences are used than when word lists are used as speech material in word identification tests. The speech intelligibility of 20 hearing-impaired speakers was measured with word identification tests using isolated words (W-22 monosyllables) and words in sentence context (CID sentences). Analysis of individual speakers' intelligibility data revealed that sentence intelligibility scores were higher than word intelligibility scores only for the better speakers and that no differences were apparent between sentence and single-word intelligibility for the poorer speakers. These findings agree with the results of research with normal speech degraded in intelligibility by noise or filtering and indicate that an interaction may exist between context and overall intelligibility in which only speech that has a certain degree of overall intelligibility may show further intelligibility improvement with increased contextual clues.

Construct validity of direct magnitude estimation and interval scaling of speech intelligibility: evidence from a study of the hearing impaired.

The appropriateness of direct magnitude estimation and interval scaling procedures for assessing the speech intelligibility of hearing-impaired adults was investigated by determining whether the continuum of the talkers' intelligibility was prothetic or metathetic. The intelligibility of 20 hearing-impaired talkers was scaled by 20 listeners using direct magnitude estimation and by 20 listeners using interval scaling. The two sets of scaling data were related in the curvilinear fashion that is typical of prothetic continua, indicating better construct validity for direct magnitude estimation than for interval scaling of speech intelligibility.