885095932

For these reasons, thc present Siudy was aimed at understanding thc effect of decodcr accuracy and knowlcdge about the topie of convcrsa-tion on the comprehension ability of thc hearing-impaired individual. The results of some successful experiments with hearing-impaired users are describcd and analyzed. The results indicate that the hearing-impaired user is able to understand the speech of thc hearing user even in the case of Iow ASR decoding accuracy (around 70%). Such an accuracy lcvcl appears to be achievable under thc real conditions de-scribed above.

6SP15. Frequency-de pendent gain control by the minima in the temporal sound envclope. Janette N. van Dijkhuizcn, Joost M. Festen, and Reinier Plomp (Dept. of Otorhinolaryngol., Free Univ. Hospital, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands)

Conventional automatic gain control acts upon the avcrage sound level, irrespective of whether therc is a speech signal or not. As a result, during periods without a speech signal, background noise is amplified to levels expericnced as “noisy" by the listener. This annoyance can be reduced by using the level of the temporal-envelope minima, rather than the average sound level, to control the gain: the information-bearing fluctuations typical for speech are preserved, whereas background noise, usually showing much less-pronounccd fluctuations, is presented at a nondisturbing level. This study investigates the effectiveness of a four-channel AGC system in which the frequency-dependent ampliflcation factor is automatically controlled by the envelope minima in the rcspec-tive frequency channcl. The reference was a condition without gain control, but with the ampliflcation in the different frcquency bands adjusted to warrant 100% intelligibility in quiet. The effect of the gain-control system on the signal appeared to be greatest in the case of stationary sounds, and smallest when a single speaker was present. Results for 10 listeners with a sensorineural hearing impairment show that, for various sounds frequently interfering in practice, with spectra that are comparable to that of the speech, the condition with gain control does not affect the speech-reception threshold in noise, but substantially reduces the subjective impression of noisiness when no speech commu-nication takes place.

, comprehension test that has been developed to address some shorteom-ings of current speech tests as measures of hearing handicap [M. J. Nilsson et a!., J. Acoust. Soc. Am. Suppl. 1 88, SI75 (1990)]. Following procedures uscd with Dutch materials (R. Plomp and A. M. Mimpen, Audiology 18, 43-52 (1979)], L^ of the scntence materials was matched and then adjusted to compensate for relative differences in difficulty between sentences. Lists of 12 sentences were created with nearly equa! distribution of phonemes. SRTs were measured using normal-hearing subjeets with unmodiffed as well as high- and low-pass filtered conditions in an effort to measure the repeatability of the lists for different signal bandwidths. Results of these tests will be presented and thc use of these materials in hearing aid research will be discussed.

6SP18. Perceptkm of /d,n,l/ by impaired- and normal-hearing listeners. Linda Kozma-Spytek, Sally G. Revoi!e (Ctr. for Auditory and Speech Sci., Gallaudet Univ., Washington, DC 20002), and James M. Pickett (Windy Hill Lab., Surry, ME 04684)

Eighteen modcrately to profoundly hearing-impaired and 11 normal-hearing listeners were studied for their use of acoustic cues for perception of /d,n,l/ in /aeOe/ tokens extracted from spoken sentences. The cues were in segments associated with the /d,n,l/ occlusions and with thc transitions of the adjacent vowcls. The segments* contribution to /d,n,l/ distinctions were examined from identiffeation tests of the /2eCae/ tokens with consonant occlusion and transition segments de-graded singly or in combination. Specifically, the consonant segments were deleted or replaced by a synthetic neutral segment; the transition segments were neutralized via substitution of adjacent pitch periods from the respective vowels. The results with normal-hearing listeners showed that the presence of any one of the consonant or transition segments supported a moderate to high level of correct /d,n,l/ identiffeation, depending on the phoneme. In contrast, the hearing-impaired listeners showed subnormal identiffeation of intervocalic /d,n,l/ which was associated with deffciencics in the use of the consonant occlusion and, gencrally, with greatly reduced use of thc vowcl transition. (Work supported by a grant from the NIDCD (NIH) and the Gallaudet Research Institute.]

6SP16. Effect of spectral envelope smearing on speech reception in noise and competing speech. Mariken ter Kcurs, Joost M. Festen, and Reinier Plomp (Dept. of Otorhinolaryngol., Free Univ. Hospital, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands)

The effect of rcduced spectral contrast on thc speech-reception threshold (SRT) for sentences in noise, and on phoneme identiffeation, was investigated with 16 normal-hearing subjeets. The SRT inereases—to about the same extent for a małe as for a female voice— as spectral energy is smeared over bandwidths exceeding thc ear’s crit-ical bandwidth. Phoncmc identiffeation shows that vowels are morę susceptible to this type of processing than consonants. Vowcls are pri-marily confused with the back vowels /d,u/, and consonants are con-fuscd wherc place of articulation is concemed. In competing speech normal-hearing subjeets show a 6-8 dB lower SRT for sentences than in steady-state noise, while sensorineurally hearing-impaired subjeets do not (J. M. Festen and R. Plomp, J. Acoust. Soc. Am. 88, 1725-1736 (1990)]. As frequency resolution may contribule to this effect, and as fluctuating mterferences of speech are very common in daily situations, the extent of the threshold differcnce between noise and speech masker was invcstigalcd for a number of smearing bandwidths.

6SP19. Methods for longitudinał measures of speech nas&lity in cochlear implant patients. M. L. Matthies, J. S. Perkell. and M. J. Williams (Res. Lab. of Electron., Rm. 36-511, MIT, Cambridge, MA ⁽ 02139)

The validity of longiludinal measures of nasality in cochlear implant patients has been examined based on acoustic spectra, sound levels, and outputs of nasal and throat accelerorneters. Speech materials consist of isolated utterances and reading passages. Preliminary observations indicate the following: The ratio of rms values of nasal and throat accel-erometer outputs (Y. Horii, Cleft Palatc J. 17, 254-261 (1980)] may be influcnced by: (a) variation in the relative levels of the two signals during the recommcnded calibration maneuver, production of a sus-tained /m/, and (b) substantial changes in SPL that accompany onset of “auditory” stimulation from a cochlear prosthesis. These observa-tions raisc uncertainty about using the throat accelerometcr output as a reference and about the sensitivity of this kind of measure to łongitudi-nal changes in nasality across experimental sessions. In addition, measures or harmonie and tbnuani amplitudes from acoustic spectra may be confounded by changes in coupling to tracheal resonances that also

accompany the activation of the prosthesis. These observations and ad-

#

ditional measures and calibration strategies will be explored further. [Work supported by NIH.]

1960

1960

6SP17. Vałldation of a speech intelligibility test using SRT for hearing aid research. Michael J. Nilsson, Jean Sullivan, and Sigfrid D. Soli (Hearing Aid Res., House Ear Inst., 2100 W. Third St., Los Angeles, CA 90057)

Speech reception thresholds (SRT) were measured in the prescncc of spectrally matched masking noise using an English language speech

J. Acoust. Soc. Am., Vol. 89, No. 4, Pt 2, April 1991 6SP20. Magnitudc estimation scaling of intelligibility by experienced and inexperienced listeners. Lee W. Ellis (Dept. of Special Education Services, The Univ. of Toledo, 2801 W. Bancroft, Toledo, OH 43615) and Donald J. Fucci (School of Hearing and Speech Sci., Lindley Hall, Ohio Univ., Athens, OH 45701)

121 st Meeting: Acoustical Society of America