Abstract: A design and implementation of multilayer perceptrons based on an architecture of multiprocessor systemon-chip is proposed in this paper.The design is highly reconfigurable,can be configured and updated on runingtime,and has the property of learning.The design proposed in this paper is verified on FPGA.The result shows that the design meets the accuracy and real-time requirements,has simple programing interface,and is easy to extend and maintain,which would benefit the applications of neural networks in computer vision and artificial intelligence.