|
Shihab Shamma
University of Maryland
Hearing as Seeing: Common principles between auditory and vision
processing
Unlike visual and tactile stimuli, auditory signals and the fundamental
cues that give rise to our percepts of timbre, pitch, and localization are
all temporal in nature. To detect and process these temporal cues, the
auditory nervous system must either possess specialized neural machinery,
or transform the responses into patterns that are spatially distributed
across its sensory epithelium and utilize more generalized networks such
as those fo und in vision. It has been common to adopt the former
hypothesis, postulating the existence of neural delay-lines, intrinsic
oscillators, and other temporal structures that facilitate processing of
the time-history of the response waveforms. However, we argue here in
favor of the opinion that the cochlea transforms sound into intricate
spatiotemporal response patterns on the auditory nerve and central
auditory stages; and that a unified computational framework, with shared
neural network architectures, exists for central auditory, visual, and
other sensory processing. Specifically, we explain how four fundamental
concepts in visual processing play analogous roles and give rise to
analogous percepts in auditory processing. These are: lateral inhibition,
multiscale cortical decomposition, temporal coincidence detection, and
spatial coincidence detection.
|
