Page 37 - Volume 12, Issue 2 - Spring 2012

Although background noise somewhat degrades phase locking in AN fibers in normal-hearing ears, the degradation is much more significant in AN fibers with noise-induced hearing loss. It is likely that this results from broadened tuning in impaired ears, which allows more noise through the auditory filter. Thus, it appears that a quantitative TFS coding deficit in the auditory periphery emerges in the presence of background noise, which may help to explain why the difficulties listeners with SNHL face are most prominent in noisy listening situations.
Beyond this quantitative TFS-coding deficit in noise, there are also qualitative degradations in TFS coding that are apparent in single AN-fiber responses to complex stimuli. Examination of the Fourier spectra of the neural correlogram functions (similar to a power-spectral density derived from an auto-correlation function), along with other system-iden- tification approaches (e.g., reverse correlation), indicate a dramatic loss of tonotopicity that can occur following noise- induced hearing loss. Whereas the normal-hearing cochlea has systematic tonotopic representation along the basilar membrane, our neural recordings from hearing-impaired chinchillas indicate that the most-prominent frequencies for which an impaired AN fiber responds to can be up to two to three octaves below the characteristic frequency (CF) corre- sponding to where the fiber innervates the cochlea. This gen- eral loss of tonotopicity (i.e., listening to information at the wrong place in the cochlea), has significant implications for almost any neural coding scheme (e.g., Oxenham et al., 2004). Similar effects have been seen previously in AN-fiber responses to vowel stimuli, where the neural coding of the second and third formants is degraded by a wide-spread response to the first formant following noise-induced hear- ing loss (Miller et al., 1997). Our results indicate that these significant degradations in the quality of temporal coding following SNHL is a much more general phenomena, which is likely to affect the neural coding of any broadband sound, particularly in background noise (also see Bidelman and Heinz, 2011; Swaminathan and Heinz, 2011).
The effects of SNHL on across-fiber temporal coding
The lack of a fundamental degradation in the ability of individual AN fibers to phase lock motivated us to also con- sider across-fiber temporal coding (i.e., spatio-temporal coding). Spatio-temporal cues, which are based on the sharp phase transitions that occur on the basilar membrane near the resonant place, have been hypothesized to be perceptu- ally relevant in a number of ways (e.g., speech, pitch, and intensity coding, tone detection in noise, binaural sound localization). Furthermore, these sharp phase transitions are predicted to be degraded by broadened tuning following SNHL. Despite the hypothesized perceptual importance of spatio-temporal cues, they have been difficult to study experimentally, e.g., due to sparse CF sampling in neuro- physiological experiments, across-neuron variability, and variability in CF estimates from the quick automated tuning- curve algorithms typically used (Chintanpalli and Heinz, 2007). Thus, the majority of studies on spatio-temporal cod- ing have been computational modeling studies and there has been a lack of experimental data characterizing the effects of
SNHL on these spatio-temporal cues, which are related to the traveling-wave delays that occur near cochlear reso- nances.
We used an innovative data-collection procedure that avoids previous experimental limitations in the study of across-fiber temporal coding (Heinz et al., 2010). The pro- cedure uses the responses of a single AN fiber to multiple frequency-shifted stimuli to predict the responses of multi- ple AN fibers with differing characteristic frequencies (CFs) to a single stimulus. The experimental benefits of this approach are that it minimizes the effects of across-neuron variability, and it provides precisely known relative CFs because they are determined based on the experimenter controlled sampling rates used. These derived population data allowed us to apply shuffled cross-correlogram analyses to quantify across-fiber correlations in both fine-structure and envelope coding, and to estimate traveling-wave delays based on characteristic delays between responses at different CFs. Similar trends in across-fiber temporal coding were observed for both the broadband noise and speech sentence stimuli we used.
Two consistent effects of SNHL were seen in all individ- ual AN fiber responses we analyzed. The cross-CF correla- tion for both TFS and ENV coding was higher following SNHL, with a broader range of CFs (width on the basilar membrane) providing correlated responses. The significance of this finding is that it suggests that SNHL produces fewer “independent” information channels for complex sounds, which would be expected to degrade perception. The second effect we observed was that the across-CF characteristic delays were reduced in AN fibers from chinchillas with noise-induced hearing loss. Although the absolute time scale of these reduced delays were quite small, they corresponded to a quarter cycle of the relevant stimulus frequency for two places on the basilar membrane separated by only a half octave. Thus, these “small” changes in across-fiber delays are potentially quite significant because they correspond to a difference between correlated and uncorrelated responses for any across-CF neural coincidence detection mechanism that may exist in the lower levels of the auditory system. Overall, these data suggest that SNHL reduces the traveling- wave delays on the basilar membrane, which degrades the sharp phase transitions that provide the spatio-temporal cues that have been hypothesized to be perceptually rele- vant. If these cues are in fact perceptually relevant, this would suggest the need for new strategies to improve hear- ing aids, which currently do not attempt to restore normal spatiotemporal responses.
Potential role of “recovered” envelopes in perceptual TFS deficits
The third hypothesis for the physiological basis of percep- tual TFS deficits that we have been exploring relates to the open question of whether it is actually possible to separate TFS and ENV components within the cochlea. Much of the evi- dence to support the idea that SNHL produces a perceptual deficit in using TFS comes from vocoded speech stimuli designed to isolate acoustic TFS and ENV (Lorenzi et al., 2006;
36 Acoustics Today, April 2012