Gerald A. Studebaker

An evaluation of some assumptions underlying the articulation index

The purposes of this investigation were to derive measurement and calculation procedures for the articulation index (AI) that were as precise and accurate as possible, and to evaluate the significance of the differences in assumptions made by early investigators. A nonsense syllable test specifically designed for this application was recorded. The implications of various assumptions utilized in different approaches to predicting speech sound identification were tested using 13 different conditions which differed with respect to filtering, level of interfering noise, or level of the speech signal. The listeners were three normal hearing individuals. A set of preferred assumptions was chosen based on a post hoc evaluation of the data. Predictions of speech sound identification were good, both for the average listener and for individual listeners.

Paired comparison judgments of relative intelligibility in noise

Normal hearing and hearing impaired subjects listened alternately to continuous discourse processed by pairs of binaural hearing aids. In each case, they were asked to judge which pair of aids reproduced speech more intelligibly. Twenty-three pairs of aids competed in three elimination tournaments and one round-robin tournament. Speech discrimination tests obtained via each aid served as the criterion measure of the success of the judgments. The results were that the paired comparison procedure produced more reliable rankings than discrimination scores particularly when the difference between aids was small. Concerning validity, the normal hearing subjects identified the aids producing the best discrimination scores very well. Also, they applied confidence ratings to their judgements that were well related to the size of the performance difference between the aids. The hearing impaired subjects also performed well generally but on occasion individuals in this group chose aids that were clearly not best for them or the group.

Evaluation of a paired comparison elimination paradigm as a method of selecting hearing aids

Continuous discourse material with a background of interfering (“Cafeteria”) noise was processed by eight different hearing aids fitted to KEMAR. Tape recordings of the processed signals were then presented in pairs to individual listeners, whose judgements of relative intelligibility were used to eliminate competing aids and select an eventual winner. Discrimination scores were also obtained from each listener, using speech discrimination tests processed under identical conditions. In general, for each individual, there was a high correlation between the aid selected by paired comparisons and that selected by the highest discrimination score. Thus paired comparisons appeared to be a reliable means of identifying hearing aids yielding maximal speech discrimination scores. However, some aspects of the results raised questions about the nature of intelligibility judgments and whether different types of test material or different characteristics of the background noise might lead to different hearing aid sel...

An analysis of articulation index theory as applied to different speech material

We have reported earlier [J. Acoust. Soc. Am. Suppl. 1 73, S102–103 (1983)] the details of an articulation index (AI) scheme that was found satisfactory for the nonsense syllable material we use. We now explore the ways of extending the scheme to include speech material of different redundancy level and different phonemic composition. Using our own data, as well as that of early AI investigators, we examine the basic assumptions of the AI theory for various speech materials. These include the properties of additivity, band independency and monotonicity, as well as the importance function, and the intelligibility‐articulation transfer function. The validity of the resulting AI scheme is discussed in terms of the magnitude of the error in speech intelligibility predictions for various speech materials. [Work supported by NINCDS.]

Effects of intensity and filtering on identification of nonsense syllables

The identification of nonsense syllables was investigated under various conditions of filtering, noise, and intensity level. The syllables were filtered and recorded simultaneously as two separate bands in a background of speech spectrum noise. The low‐pass band (200–400 Hz) was recorded at a nominal signal to noise ratio of 20 dB, the high‐pass band (1–5 kHz) at 7 dB. The low‐pass band was presented at 60, 80, and 96 dB SPL, the high‐pass band at 60 and 80 dB SPL (nominal levels represent coupler level prior to filterings. Each band was presented alone and in every possible combination of intensity level with the other band to four normal hearing subjects. The best performance resulted when both bands were at 60 dB. Increasing the level of the high‐pass band by 20 dB and of the low‐pass band by 20 and 36 dB relative to the best condition resulted in a reduction in performance. Poorest performance resulted when the low‐pass band was at the highest level (96 dB) and the high‐pass band was at the lowest lev...

An evaluation of certain assumptions underlying articulation index testing

The implications of various assumptions utilized in different approaches to the articulation predictions were tested utilizing several filtered speech, interfering noise and low level listening conditions. The monotonicity principle, basic to the AI theory, was examined and found valid when certain plausible adjustments were made in the speech signal and auditory sensitivity levels. The possibility of different intelligibility‐articulation index transfer functions for amplitude distortion and noise interference was evaluated. An analysis of prediction results obtained by using different calculation techniques revealed that they did not agree. The band independency principle was investigated, both as related to the general validity of the AI concept and also as related to the test materials and calculation procedures used. Variations of certain assumptions within a given calculation technique (e.g., real peaks versus 12 dB across all frequencies) were found to have only minor effects on the accuracy of pre...

Directivity of the human vocal source

The directivity affects of the voices of one male and one female talker were observed in several environments including an anechoic room. Measurements were made at all frequencies between about 200 Hz and 10 kHz at 45°, 90°, 135°, 180°, and 270° azimuths. All measurements were made in the horizontal plane at the level of the mouth opening and at one meter from the center of the head. The results agreed well with the only other known comparable studies [Dunn and Farnsworth, 1939; Moreno and Pfretzschner, 1979] and also agree well with results obtained by Olson (1972) who designed and used a simplified manikin source. Some loudspeaker types produce results very different from those of human talkers. A search is underway to determine whether there are any widely available or easily created loudspeaker configurations that reasonably closely replicate the observed directivity of human sources. [Work supported by NINCDS.]

A nonsense syllable test designed for articulation index testing

The band‐importance function developed by Bell Laboratories is not directly applicable to any of the available recorded speech tests. For this reason we developed a test which is like that used by the Bell Laboratories but unlike other available tests in that (1) each CV or VC is a different utterance (2) the phonemes appear with uniform frequency. The same phonemes used by the Bell Laboratories were used. A “test” consists of utterances by one male and one female talker each speaking 152 CV and VCs following a carrier phrase. The frequency spectrum of this material, as well as the procedure used for its estimate, will be described. Practice effects were studied at different magnitudes of distortion. The importance function which is presumably applicable to this material was found valid within the error of measurement for the filtered speech articulation test results. The error in predicting the articulation score will be related to the magnitude of the intrinsic variability of articulation testing. [Wor...

On improving articulation index predictions of speech sound identification performance by the hearing impaired

Predictions of speech sound identification performance by hearing impaired individuals were made using an articulation index (AI). The AI scheme used was one that had earlier been found accurate for normal hearing listeners. It was expected that some sensorineural hearing impaired individuals would exhibit reduced suprathreshold speech processing capacity, and that therefore, the AI procedure would need to be modified to accomodate those effects. To this end we measured several psychoacoustical variables (critical bands, upward spread of masking, temporal masking, tuning curves, etc.) that might be incorporated into the AI formulas to account for the degraded performance. Results for each of the individuals tested will be discussed. [Work supported by NINCDS.]

Paired comparison judgements of relative speech intelligibility

The reliability, validity, and sensitivity of paired comparison judgements of speech intelligibility, organized into round robin and elimination tournaments, were studied. Normal hearing and hearing impaired subjects listened to paired samples of speech that were processed with two levels of background noise by 23 hearing aids in two replications. Monosyllabic word intelligibility scores were obtained under each condition in both replications. The results indicated that this paired comparison procedure was a valid, reliable and reasonably sensitive means of evaluating the relative intelligibility of hearing aid processed signals. For example, subjects chose the aid with the best, or second best average discrimination score in 94% of the elimination tournaments. Regarding reliability, the aid that was chosen first on the first trial was the first or second choice on the second trial more than 90% of the time at the O dB S/N ratio. The sensitivity of the procedure was such that at the O dB S/N ratio, the be...