Cao Kl

Mandarin tone recognition in cochlear-implant subjects

This study examined tone recognition in five cochlear-implant subjects who were native speakers of Mandarin and used a Nucleus-22 device. Psychophysical experiments were conduced to measure rate discrimination in individual electrodes from the most apical to the most basal electrodes. The rate range was from 100 to 200 Hz, which corresponded to the range of variation in fundamental frequency for the tonal tokens used in this study. Speech recognition experiments were also conducted to measure tone recognition as function of the number of electrodes from a 1-electrode map to a 20-electrode map. Large individual variability was observed for both rate discrimination and tone recognition result: Average rate discrimination ranged between 0.2 and 1.2 (Webers fraction) whereas tone recognition ranged between 30% and 70% correct. A highly significant correlation was found between rate discrimination and tone recognition with the 20-electrode map, but a non-significant correlation was observed with the 1-electrode map due to a floor effect in tone recognition. The present result supports the hypothesis that both spectral and temporal cues contribute to tone recognition. In addition, the present result shows that current cochlear-implant subjects produced significantly lower performance than acoustic simulations in normal-hearing subjects, suggesting that neither temporal nor spectral cues have been adequately and appropriately extracted and encoded in current cochlear implants. New designs are discussed to improve tone recognition in cochlear implant subjects.

Psychophysical performance and Mandarin tone recognition in noise by cochlear implant users.

Objective: The present study was aimed to examine the relationship between psychophysical performance in temporal and spectral resolution and Mandarin tone recognition in noise by cochlear-implant (CI) listeners. Design: Seventeen Nucleus-24 implant users, 10 postlingually deafened and 7 prelingually deafened, participated in the experiments. A 3-interval, forced-choice procedure was used to measure gap detection and pure-tone frequency discrimination at 250 to 4,000 Hz in octave steps. A 4-alternative forced-choice procedure was used to measure Mandarin tone recognition in quiet and in noise. Signal-to-noise ratios (SNRs) varied from +10 to –10 dB. All stimuli were delivered to the clinical processor via a speaker in a sound free field. The obtained data were compared to data collected from normal-hearing control subjects, as well as cochlear-implant users who performed similar tasks using single-electrode stimulation via a research interface. Results: Postlingually-deafened CI subjects generally performed better than prelingually-deafened subjects. The average gap detection threshold was 30 ms with a range from 4 to 128 ms. The average frequency difference limen was 100 Hz with a range from 12 to 192 Hz, regardless of the standard frequency. The average tone recognition was 80% correct in quiet, which dropped to 55% at +10 dB SNR and essentially chance performance at –5 dB SNR. In comparison, the normal-hearing control subjects maintained essentially perfect performance over this SNR range. Only frequency discrimination at 1,000 Hz was significantly correlated with tone recognition in quiet but all psychophysical measures were correlated to tone recognition in noise. Conclusions: The present result suggests that the CI users can rely on either temporal or spectral cues to perform tone recognition in quiet, but need both cues for tone recognition in noise. Future CI processors need to extract and encode these acoustic cues to achieve better performance in tone perception and production.