Design and Implementation of a Lightweight Artificial Intelligence Model for Intelligent Terminals
Main Article Content
Abstract
This paper designs and implements a lightweight artificial intelligence model for intelligent terminals. First, the hardware characteristics of smartphones, IoT terminals, and wearable devices are analyzed, together with the real-time, low-power, and storage-constrained requirements of terminal-side AI applications. Second, the design objectives and constraints of the lightweight model are clarified, and an overall architecture named LightNet is constructed. The model optimizes the feature extraction and inference decision modules using depthwise separable convolution, a bottleneck structure, attention enhancement, operator fusion, pruning, and 8-bit quantization. The model is implemented and deployed using the PyTorch framework, and terminal debugging is completed on mainstream intelligent devices. Experimental results show that the proposed model reduces the number of parameters by up to 96.9% compared with VGG16, improves inference speed by 77.5%, and reduces power consumption by 64.3% while maintaining a classification accuracy of 92.3%. Because intelligent terminals often operate through antenna-enabled sensing and wireless electromagnetic propagation environments, the model is applicable to edge AI scenarios requiring low-latency, energy-efficient, and electromagnetic-compatible deployment.
Downloads
Article Details
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
References
A. Gholami, S. Kim, Z. Dong, et al., “A Survey of Quantization Methods for Efficient Neural Network Inference,” arXiv, 2021, doi: 10.48550/arXiv.2103.13630.
Z. Qin, “Research on Optimization of Lightweight Convolutional Neural Network Models for Computer Vision,” National University of Defense Technology, 2018, doi: 10.27052/d.cnki.gzjgu.2018.001449.
G. Howard A, M. Zhu, B. Chen, et al., “MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” arXiv, 2017, doi: 10.48550/arXiv.1704.04861.
M. Sandler, A. Howard, M. Zhu, et al., “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” IEEE, 2018, doi: 10.1109/CVPR.2018.00474.
X. Zhang, X. Zhou, M. Lin, et al., “ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices,” IEEE, 2018, doi: 10.1109/CVPR.2018.00716.
H. Tampubolon, A. Setyoko, and F. Purnamasari, “SNPE-SRGAN: Lightweight Generative Adversarial Networks for Single-Image Super-Resolution on Mobile Using SNPE Framework,” Journal of Physics: Conference Series, vol. 1898, no. 1, Art. no. 012038, 2021, doi: 10.1088/1742-6596/1898/1/012038.
K. Han, Y. Wang, Q. Tian, et al., “GhostNet: More Features From Cheap Operations,” IEEE, 2020, doi: 10.1109/CVPR42600.2020.00165.
G. Zhang, G. Xie, W. Yan, et al., “Paddle Lite on Zephyr: Deploying AI Models in RTOS for Inference Acceleration,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. (1), pp. 45, 2026, doi: 10.1109/TCAD.2025.3581867.
X. Xing, Z. Liu, S. Xiao, et al., “EfficientLLM: Scalable Pruning-Aware Pretraining for Architecture-Agnostic Edge Language Models,” 2025.
Z. Liu, J. Li, Z. Shen, et al., “Learning Efficient Convolutional Networks through Network Slimming,” IEEE, 2017, doi: 10.1109/ICCV.2017.298.
S. Tan Y, T. Li, and S. Zhang Y, “Research on Generation and Deployment of Neural Network Models for Edge Intelligence,” Computer Engineering, vol. 50, no. 8, pp. 1-12, 2024, doi: 10.19678/j.issn.1000-3428.0068554.
C. Alippi, S. Disabato, and M. Roveri, “Moving Convolutional Neural Networks to Embedded Systems: The AlexNet and VGG-16 Case,” ACM, pp. 212-223, 2018, doi: 10.1109/IPSN.2018.00049.
S. Li, L. Wang, J. Li, et al., “Image Classification Algorithm Based on Improved AlexNet,” Journal of Physics Conference Series, vol. 1813, no. 1, Art. no. 012051, 2021, doi: 10.1088/1742-6596/1813/1/012051.
C. Parlak and Y. Altun, “A Quest for Formant-Based Compact Nonuniform Trapezoidal Filter Banks for Speech Processing with VGG16,” Circuits, Systems, and Signal Processing, 2026, doi: 10.1007/s00034-024-02794-z.
D. Liu, Y. Zhu, Z. Liu, et al., “A survey of model compression techniques: past, present, and future,” Frontiers in Robotics & AI, 2025, doi: 10.3389/frobt.2025.1518965.
N. Lai, A. Dewi D, S. Maidin S, et al., “A comprehensive review of lightweight deep learning models for edge computing with future directions,” Discover Computing, pp. 29(1), 2026, doi: 10.1007/s10791-026-10021-3.
A. Musa, A. Kakudi H, M. Hassan, et al., “Lightweight Deep Learning Models For Edge Devices—A Survey,” International Journal of Computer Information Systems and Industrial Management Applications, vol. 17, pp. 18, 2025, doi: 10.70917/2025014.
V. Sze, H. Chen Y, J. Yang T, et al., “Efficient Processing of Deep Neural Networks: A Tutorial and Survey,” Proceedings of the IEEE, Art. no. 105(12), 2017, doi: 10.1109/JPROC.2017.2761740.