VLSI Architecture for Energy-Efficient Convolutional Neural Networks in Embedded Image Recognition Systems
DOI:
https://doi.org/10.29284/ijasis.12.1.2026.46-62Keywords:
VLSI architecture; convolutional neural network (CNN); energy efficiency; approximate computing; embedded vision; reconfigurable processing elements; MAC optimization; quantization; low-power design; edge intelligence.Abstract
The integration of Convolutional Neural Networks (CNNs) in embedded vision systems demands a careful balance between computational efficiency, power consumption, and on-chip memory utilization. This paper presents a novel energy-efficient VLSI architecture optimized for CNN-based image recognition in resource-constrained embedded platforms. The proposed design employs approximate arithmetic computing, clock gating, and data reuse optimization to achieve significant energy savings without compromising recognition accuracy. The architecture introduces a reconfigurable processing element (RPE) array with shared weight buffers and hierarchical memory control to minimize data movement between layers. A pipelined convolutional engine combined with quantized 8-bit fixed-point operations reduces both dynamic power and hardware complexity. Additionally, a parallel multiplier–accumulator (MAC) optimization is incorporated using a modified Booth algorithm for faster convolution and reduced latency. Implemented in 45 nm CMOS technology using Cadence and Synopsys Design Compiler, the proposed architecture achieves a 35% reduction in power, 28% lower area utilization, and 40% improvement in throughput compared to conventional CNN accelerators. When deployed on embedded benchmarks such as CIFAR-10 and MNIST, the system attains a classification accuracy of 98.3% with a peak energy efficiency of 1.26 TOPS/W. The design demonstrates excellent scalability and suitability for edge-based visual intelligence applications such as drones, surveillance, and IoT-enabled smart devices.
References
1. Wang, Y., Lin, J. and Wang, Z., 2017. An energy-efficient architecture for binary weight convolutional neural networks. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 26(2), pp.280-293.
2. Park, S., Hong, I., Park, J. and Yoo, H.J., 2016. An energy-efficient embedded deep neural network processor for high speed visual attention in mobile vision recognition SoC. IEEE Journal of Solid-State Circuits, 51(10), pp.2380-2388.
3. Kang, M., Gonugondla, S.K., Keel, M.S. and Shanbhag, N.R., 2015, April. An energy-efficient memory-based high-throughput VLSI architecture for convolutional networks. In 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (pp. 1037-1041). IEEE.
4. K.Kalpana., Dr.B.Paulchamy.,2022, "A novel design of nano router with high-speed crossbar scheduler for digital systems in QCA paradigm", Circuit World, Vol. 48 No. 4, pp. 464-478.
5. Kumar, V., Baskar, D., Arunsi, M., Jenova, R. and Majid, M.A., 2019. VLSI design and implementation of High-performance Binary-weighted convolutional artificial neural networks for embedded vision based Internet of Things (IoT). Procedia Computer Science, 163, pp.639-647.
6. Kang, M., Keel, M.S., Shanbhag, N.R., Eilert, S. and Curewitz, K., 2014, May. An energy-efficient VLSI architecture for pattern recognition via deep embedding of computation in SRAM. In 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (pp. 8326-8330). IEEE.
7. Deepika, S. and Arunachalam, V., 2023. Design of high performance and energy efficient convolution array for convolution neural network-based image inference engine. Engineering Applications of Artificial Intelligence, 126, p.106953.
8. Li, J., Un, K.F., Yu, W.H., Mak, P.I. and Martins, R.P., 2021. An FPGA-based energy-efficient reconfigurable convolutional neural network accelerator for object recognition applications. IEEE Transactions on Circuits and Systems II: Express Briefs, 68(9), pp.3143-3147.
9. Reddy, P.S.S. and Reddy, P.R.S., 2024, November. Energy-Efficient VLSI Architecture for Real-Time Retinal Disease Detection using Deep Learning. In 2024 5th International Conference on Data Intelligence and Cognitive Informatics (ICDICI) (pp. 1120-1125). IEEE.
10. Biswas, A. and Chandrakasan, A.P., 2018, February. Conv-RAM: An energy-efficient SRAM with embedded convolution computation for low-power CNN-based machine learning applications. In 2018 IEEE International Solid-State Circuits Conference-(ISSCC) (pp. 488-490). IEEE.
11. Ravi, S., Sureshbabu, J., Navaneethan, S., Chadge, R., Al-Qawasmi, K. and Kumar, B.R., 2025, July. Energy-Efficient VLSI Architectures for On-Device Deep Learning in Real-Time Vision Applications. In 2025 6th International Conference on Data Intelligence and Cognitive Informatics (ICDICI) (pp. 1024-1029). IEEE.
12. Chen, Y.H., Krishna, T., Emer, J.S. and Sze, V., 2016. Eyeriss: An energy-efficient reconfigurable accelerator for deep convolutional neural networks. IEEE journal of solid-state circuits, 52(1), pp.127-138.
13. Deepika, S. and Arunachalam, V., 2025. An energy efficient processor array and memory controller for accurate processing of convolutional neural network-based inference engines. Scientific Reports, 15(1), p.39693.
14. Manikandababu, C.S., Jagadeeswari, M. and Irfan, H.M., 2023, November. Low-Power VLSI Design for Image Analysis in Embedded Vision Systems. In 2023 International Conference on Sustainable Communication Networks and Application (ICSCNA) (pp. 838-844). IEEE.
15. Narayanan, K.L., Al-Zubi, A., Bostani, A., Norboyev, S., Buyukbicakci, E., Niphadkar, C. and Omprakash, S., 2025. Energy-Aware VLSI Architectures for Intelligent Signal Processing in Neural Computing and Next-Generation IoT–Robotics Integration. Journal of VLSI Circuits and Systems, 7(1), pp.253-261.
16. Lee, J., Kim, C., Kang, S., Shin, D., Kim, S. and Yoo, H.J., 2018. UNPU: An energy-efficient deep neural network accelerator with fully variable weight bit precision. IEEE Journal of Solid-State Circuits, 54(1), pp.173-185.
17. Dhanasekar J, Sudha VK. ,2024, Relative Feature Classification Method for Diabetic Retinopathy Detection from Scanned Eye Images, Indian Journal of Science and Technology. 17(31): 3244-3252.
18. K. Sivakami, S. Arunkumar, M. P. Usharani, J. Dhanasekar, S. Kokilamani and N. Revathi, "Data Clustering and Routing using Improved Water Wave Optimization in Wireless Sensor Network," 2024 5th International Conference on Communication, Computing & Industry 6.0 (C2I6), Bengaluru, India, 2024, pp. 1-6.
19. Duan, Y., Li, S., Zhang, R., Wang, Q., Chen, J. and Sobelman, G.E., 2018, November. Energy-efficient architecture for FPGA-based deep convolutional neural networks with binary weights. In 2018 IEEE 23rd International Conference on Digital Signal Processing (DSP) (pp. 1-5). IEEE.
20. Kim, S., Jo, J. and Park, I.C., 2021. Hybrid convolution architecture for energy-efficient deep neural network processing. IEEE Transactions on Circuits and Systems I: Regular Papers, 68(5), pp.2017-2029.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
