Encoding the Acoustic Spectrum in the Spatio-Temporal Responses of the Auditory-Nerve

Abstract

Acoustical spectral features play an important role in the auditory perception and recognition of speech and complex sounds (Fant, 1973). These features are initially derived in the peripheral auditory system through the fine frequency selectivity of the basilar membrane (BM). Given the remarkable stability and robustness of the acoustic percepts, the underlying spectral cues must be preserved in the responses of the auditory-nerve over large ranges of intensity and noise conditions. Experimental evidence from single unit studies, however, has not been easy to interpret because several nonlinear phenomena of cochlear function conspire to alter drastically the apparent nature of nerve responses for different stimulus conditions (Sachs and Young, 1980). Consequently, several response measures have been proposed to detect and extract reliably the encoded spectral parameters. They may be organized along a continuum between two extremes: (1) Purely spatial measures that utilize only the spatial profile of the average rate of response along the tonotopically organized nerve-fiber array (e.g. place-code theory) (Sachs and Young, 1979). (2) Temporal periodicity measures that dispose of the tonotopic axis, using instead the periodicities in the response (phase locking) as measures of the stimulus spectral content (e.g. Dominant Frequency algorithm) (Sinex and Geisler, 1983).