A Biological Approach to Pattern Recognition

This paper describes a layer structured system suitable for pattern recognition which operates similar to the afferent nervous system of vertebrates. The ``system theory of homogeneous layers'' has been developed to describe signal transmission and signal processing between neuronal layers. Feature extraction in the sense of spatial filtering is performed by such a layered system with a few hierarchical stages. The last stage contains adaptive coupling which is adjusted by a learning process. The system has been simulated with a computer and parts of it with a coherent light arrangement. Its performance im recognizing handprinted characters (alphanumerics) is highly satisfactory and corresponds approximately to the human capability for this task.