A Novel Solid Neuron-Network Chip Based on Both Biological and Artificial Neural Network Theories

Built on the theories of biological neural network, artificial neural network methods have shown many significant advantages. However, the memory space in an artificial neural chip for storing all connection weights of the neuron-units is extremely large and it increases exponentially with the number of neuron-dentrites. Those result in high complexity for design of the algorithms and hardware. In this paper, we propose a novel solid neuronnetwork chip based on both biological and artificial neural network theories, combining semiconductor integrated circuits and biological neurons together on a single silicon wafer for signal processing. With a neuro-electronic interaction structure, the chip has exhibited more intelligent capabilities for fuzzy control, speech or pattern recognition as compared with conventional ways.