Tactile Pattern Recognition

Abstract

The broad aim of our research is to understand tactile pattern recognition. Pattern recognition in all of the sensory systems occurs so rapidly relative to neuronal processing times that elaborate, serial algorithms are ruled out. Pattern recognition must consist of some form of matching between incoming neural images and images stored in memory that is accomplished in a time frame amounting to a small number of neuronal steps. It is possible that the final images, which enter memory, are simple isomorphic representations of the stimulus, but that would require that every recognizable pattern be stored at every possible location, size and orientation for rapid matching. It seems more likely that the images being matched are in some other (non-isomorphic) form that allows rapid matching regardless of size, location and orientation. In both the visual and somatosensory systems the topological organization of the receptor sheet dictates that the primary neural representation of a spatial stimulus is an isomorphic spatial pattern of neural activity. One task of the ascending pathways within these sensory systems must be to transform neural images from this primary isomorphic form to a form that allows rapid recognition, perception, and association. We believe that the key to pattern recognition in the somatosensory system is to understand the central mechanisms of transformation and representation.