Bionic pulse neural network visual identification system based on FPGA

Abstract

The invention relates to a bionic pulse neural network visual identification system based on an FPGA. The system is characterized in that the system comprises an FPGA development board and an upper computer which are connected with each other, wherein the FPGA development board comprises a preprocessing module, a feature extraction module and a pulse neural network model module, and the preprocessing module, the feature extraction module and the pulse neural network module are all compiled by adopting a Verilog HDL language and compiled and downloaded to the FPGA development board by a QUARTUS II; and the upper computer is connected with the FPGA development board through the USB interface module so as to communicate with the FPGA chip through the I / O port. External environment information is collected through a camera, image feature signals are obtained through processing of a preprocessing module and a feature extraction module, and pulse signals are generated after encoding of a pulse encoding module and input into a pulse neural network module.

Description

technical field [0001] The invention relates to biomedical engineering technology and image processing technology, in particular to an FPGA-based bionic pulse neural network visual recognition system. Background technique [0002] The human brain can perform fast and reliable object recognition with extremely low power consumption, which is more powerful than current computer vision systems. Neurophysiological studies have shown that the ventral visual pathway with a hierarchical structure in the human brain is closely related to object recognition, and it includes four areas: V1, V2 / V4 and IT. Although traditional artificial neural networks have achieved remarkable results in solving pattern recognition and classification tasks, they are only structural simulations, which are completely different from the brain in terms of neuron models and information processing mechanisms. As the third-generation artificial neural network, the spiking neural network has attracted extensi...

AI Technical Summary

Problems solved by technology

[0005] The existing technology is still in its infancy, so there are still the following disadvantages: there is no FPGA-based bionic spiking neural network visual recognition platform with complete functions; the hardware simulation spiking neural network implemented by FPGA is small in scale, and the processing speed of traditional visual recognition systems is relatively small. It is far from meeting the needs of realization; the human-machine interface is not yet perfect, and it is impossible to perform real-time control and data analysis

Images