Application and Implementation of Convolutional Neural Network Accelerator Based on FPGA in Environmental Sound Classification

With the development of artificial intelligence, convolutional neural network (CNN) has achieved great success in various learning tasks, and has been applied to environmental sound classification tasks, achieving a high classification accuracy. With the development of convolutional neural network in the application of environmental sound classification, the design of hardware accelerator for convolutional neural network has become a hot topic. Compared with CPU, GPU and ASIC, FPGA has the unique advantages of low power consumption and high energy efficiency in accelerating deep learning, and has become the preferred hardware platform for neural network hardware implementation. In this paper, a convolution neural network accelerator based on FPGA is proposed, which can realize convolution, tanh nonlinear activation, average pooling and other operations. The hardware accelerator can realize the inference process of environmental sound classification, and achieve high energy efficiency and low resource consumption while having a high classification accuracy. The network was tested on Xilinx Virtex VC-707 FPGA board. The experimental results show that its energy efficiency is 547.60 and 377.36 times higher than that of CPU and GPU respectively.