Page 38 - Volume 12, Issue 2 - Spring 2012

Fig. 3. Theoretical framework for evaluating the neural basis for the perceptual salience of acoustic temporal fine-structure (TFS) and envelope (ENV) cues. Acoustic TFS can produce recovered ENV cues at the output of the cochlea in addition to true neural TFS cues. Recovered envelopes arise due to narrowband cochlear filtering (*; Ghitza, 2001), and therefore are reduced or not present in lis- teners with sensorineural hearing loss or in cochlear-implant patients. Thus, cau- tion is required in applying perceptual results from normal-hearing listeners to the design of auditory prostheses. (Figure is taken from Heinz and Swaminathan (2009), with permission from Springer-Verlag and the Association for Research in Otolaryngology.)
Hopkins et al., 2008). The physiological basis for these per- ceptual results is difficult to evaluate because narrowband cochlear filtering limits the ability to isolate fine-structure and envelope at the output of the cochlea. Specifically, inter- pretation of perceptual results using these specialized acoustic stimuli (e.g., TFS speech) relies on the assumption that envelope and fine-structure can be isolated at the output of the cochlea. However, narrowband cochlear filtering imposes constraints on isolating a sound’s fine-structure from its envelope (Ghitza, 2001), as predicted by several sig- nal processing theorems (Voelcker, 1966; Rice, 1973; Logan, 1977). Thus, proper interpretation of the perceptual salience of TFS cues must include consideration of the fact that acoustic TFS not only produces true neural TFS cues, but also recovered envelope cues (see Fig. 3). In other words, acoustic TFS cues can produce useful temporal coding at the output of the cochlea in two ways: (1) AN responses syn- chronized to stimulus fine structure itself (“true TFS”), and (2) AN responses synchronized to stimulus envelope (i.e., “recovered envelopes,” which are created by narrowband cochlear filters). These recovered neural ENV cues arise when broadband TFS signals are passed through narrow- band cochlear filters, which perform an FM-to-AM conver- sion such that their outputs “recover” ENV patterns that are correlated with the ENV patterns from the original stimulus. It is then difficult to interpret whether observed perceptual effects related to acoustic TFS cues are truly based on neural TFS coding, or perhaps based on neural ENV cues. Of course, the answer to this question is critical for the transla- tional implications of studies demonstrating perceptual TFS deficits.
We have been exploring this issue through a coordinated use of neurophysiological, perceptual, and computational- modeling approaches. Based on previous perceptual studies showing that recovered envelopes can contribute to the per- ception of TFS cues for speech in quiet conditions (Zeng et al., 2004; Gilbert and Lorenzi, 2006), we collected neural data
Fig. 4. A computational auditory-nerve (AN) model predicts spike-train responses of a single AN fiber to an acoustic stimulus. The characteristic frequency (CF) of the fiber, as well as the degree of inner-hair-cell (CIHC) and outer-hair-cell (COHC) dysfunction can be specified. This model has been tested against physiological data from animals with normal hearing or with noise-induced hearing loss. We used this model to predict neural temporal fine-structure TFS and ENV coding in response to the same vocod- ed speech stimuli for which consonant identification was measured in human listeners. Results suggest that recovered neural ENV cues from acoustic TFS cues provide impor- tant information for speech perception in noise. Predictions with OHC dysfunction suggest that these recovered ENV cues are reduced following sensorineural hearing loss (SNHL). (Figure modified with permission from Zilany and Bruce (2006). Copyright 2006, American Institute of Physics.)
Physiological Correlates of Perceptual Deficits 37