Recognition of Sounds by Cochlear Patterns

Abstract

An electrical analog of the human ear has been developed to provide real-time cochlear patterns of subjective loudness along the basilar membrane. Resulting spatial patterns may be analogous to those found in the auditory centers of the central nervous system. It is hypothesized that cochlear pattern shapes are of primary importance in the recognition of sound. It is further hypothesized that the cochlea performs a partial analysis of the sound and that the higher analysis centers of the central nervous system perform additional analyses. Concepts and processes of analysis and recognition are developed. Analysis is discussed from the viewpoint of an information mapping process in a multidimensional space. Recognition is discussed as a process of locating unknown points (patterns) in multidimensional space by relative measures to known points. An experiment is described which demonstrates the similarity in recognition between the human and the analog using a recognition function based on cross correlation.