Design And Implementation Of Neural Network Processing System Based On ARM And FPGA

The implementation study and practice of artificial neural network has become a hot topic in the field of intelligent control since the ANN theory was proposed. Implement way of artificial neural networks is gradully changing from rely on computer simulation to hardware; this is because the hardware way is not only more efficient, but also a combination of a computer system with a microprocessor for embedded system design. Traditional neural network processor relys solely on computer system or hardware, and not be able to combine the advantages of both, but embedded system and equipment can achieve this. ARM microprocessor with powerful extensions and control functions, while FPGA with a fast arithmetic processing functions; artificial neural network processor based on both can be highly reliable and efficient.Through the study of the structural features and algorithms of artificial neural network model, and based on ARM and FPGA features, the neural network operating principle and processes are studied; and a neural network processor data storage solutions and a communication scheme for ARM microprocessor and FPGA neural network processor are designed. And studied the linux kernel tailoring, customization and migration, as well as under linux system DMA data transfer and interrupt control design. In this paper, the design and implementation of embedded neural network processor provides a reference value of the theoretical and technical support.