심층신경망의 스마트응용을 위한 온칩 시스템 = An On-a-Chip System for Smart Applications based on DNN (Deep Neural Network)

다국어 초록 (Multilingual Abstract)

Although Deep Neural Network (DNN) shows excellent performance in pattern recognition such as image processing and voice recognition, it has the disadvantage of requiring intensive computing powers for real-time processing. This paper presents a programmable On-a-chip hardware for embedded smart applications based on Deep Neural Network (DNN) computations. To implement and test the proposed on-chip system, an FPGA-based development platform is produced and used. Since the main processor and related hardwares can be implemented in the same FPGA using the platform, all hardware blocks can be integrated and verified easily. A operation program is provided for On-a-chip system core initialization, operation, and software development environment. The effectiveness of the proposed On-a-chip system is shown through an experiment implementing an actual application system based on DNNs.

참고문헌 (Reference)

1 Albawi, S., "Understanding of a convolutional neural network" 1-6, 2017

2 LeCun, Y., "Mnist handwritten digit database"

3 Gaisler Research, "LEON2 processor user’s manua"

4 Alex Krizhevsky, "ImageNet - Classification with Deep Convolutional Neural Networks" 1097-1105, 2011

5 M. Shabiul, "Design and Implementation of Discrete Cosine Transform Chip for Digital Comsumer Products" 52 (52): 998-1003, 2006

6 P. G. D. Valle, "Application of FPGA Emulation to SoC Floorplan and Packaging Exploration" 2007

7 Chen, Y., "A survey of accelerator architectures for deep neural net works" 6 (6): 264-274, 2020

8 이봉규, "A programmable Soc for Var ious Image Applications Based on Mobile Devices" 한국멀티미디어학회 17 (17): 324-332, 2014

9 G. Kwon, "A Study of Edge AI Hardware for Real-Time Inference of Convolutional Neural Network Model Using FPGA" 26 (26): 273-276, 2022

10 Lulin Chen, "A CNN(Convolutional Neural Network) hardware implementation"