Information processing in biology-inspired pulse coded neural networks

This paper briefly reviews the characteristic properties of biological neural systems and describes the recent development of a biology-inspired pulse processing neural network (BPN) hardware. The BPN employs pulse processing, EPSP, IPSP, membrane potential, refractory period, adaptive synaptic weights, and adaptive time delays, as well as asynchronous pulse train communication in real time. Information is coded not only in the impulse rate but also in the time of occurrence of each individual impulse. A typical BPN application for adaptive recognition and subsequent tracking of discrete time event patterns using adaptive time delays and representation of the "learning rule" by means of special neurons and pre-synaptic events is presented. BPN systems may be particularly useful in the emerging field of "neurotechnology", which deals with the repair or partial compensation of functional deficits of the human nervous system by means of coupling prosthetic devices via multicontact single axon interfaces and BPNs with specific parts of the nervous system.