| Abstrakt: |
Psychophysical models often begin with a bank of filters and assume that the energy in the filter outputs is the basis for coding complex sounds. The filter bandwidths are estimated based on masked detection thresholds and are referred to as critical bands. While this conceptual model structure can explain many psychophysical phenomena, it also fails for many tasks. We propose a different conceptual model: a bank of peripheral filters, each followed by a saturating nonlinearity that represents transduction by inner hair cells (IHCs). The metric of interest is the amplitude of low-frequency fluctuations in the hair-cell responses. Low-frequency fluctuations in the IHC voltage drive fluctuations in the auditory-nerve (AN) responses, referred to here as "neural fluctuations" (NF). NFs are a temporal response feature distinct from phase-locking to temporal fine structure, and they influence AN responses across a wide range of frequencies. The saturating nonlinearity that follows each peripheral filter results in neural fluctuations that differ markedly from both the envelope of the stimulus and the envelope of a bandpass-filtered stimulus representing the basilar membrane response. A key difference between neural fluctuations and stimulus-related envelopes occurs when IHCs tuned near a spectral peak are "captured" (or dominated) by the peak in the stimulus spectrum (Deng and Geisler, 1987; Zilany and Bruce, 2007). A peripheral channel that is captured has relatively small fluctuations in the response, whereas channels that are not captured can have strongly fluctuating responses, due to beating between multiple stimulus components. The profile of neural fluctuation amplitudes along the characteristic-frequency (CF) axis encodes key features of the stimulus spectrum. For example, in the case of masking paradigms, a local decrease in the neural-fluctuation amplitude profile encodes the presence and frequency of the target tone. An understanding of the role of the saturating IHC transduction nonlinearity in the proposed model is facilitated by a computational model for the periphery that includes the interaction between the saturating nonlinearity and compressive cochlear gain. We will demonstrate that neural-fluctuation profiles in response to psychophysical masking paradigms can predict performance in several tasks for which the classical critical-band model fails (as well as those for which it succeeds.) Our goal is to determine whether a single model can predict results for several masked-detection tasks, including band-widening paradigms, with or without roving-levels, and co-modulation masking release. The profile of neural fluctuations along the tonotopic axis is of interest because most midbrain neurons are excited or suppressed over a range of amplitude-modulation frequencies (Kim et al., 2020). The fluctuation profiles that are set up in the periphery are thus transformed into rate profiles at the level of the auditory midbrain. Rate profiles across populations of model midbrain neurons were the basis for the psychophysical predictions described below. Finally, the degree of contrast in the neural-fluctuation profiles is reduced by sensorineural hearing loss, suggesting that fluctuation profiles may provide insight into deficits associated with hearing loss. [ABSTRACT FROM AUTHOR] |
