Design of handwritten digit recognition system based on hummingbird E203 RISC-V processor
Handwritten digit recognition is a classic problem in the field of computer vision,playing an important role in areas such as license plate recognition and optical character recognition.Deploying high-performance handwritten digit recognition sys-tems in embedded devices,due to the constraint of ARM and X86 architecture,the system's computing power,cost,power con-sumption and other indicators are not ideal.The RISC-V architecture has advantages such as open source,simplicity,strong scal-ability,and well-organized instruction encoding,and has received high praise in the industry in recent years.This article optimizes the open-source Hummingbird E203 RISC-V processor and adds a convolutional neural network coprocessor unit to complete the recognition of handwritten digits.The test results show that when the system operates at a frequency of 25MHz,the convolutional neu-ral network coprocessor designed with the Hummingbird E203 RISC-V processor takes an average recognition time of 1ms for hand-written digit recognition.The average frame rate for processing video stream data is 912 frames,with an accuracy rate of 98%,which confirms the feasibility of this system and demonstrates the superiority of RISC-V over ARM and X86 architecture processors.
万方数据
