Image compression approaches using implicit neural representation (INR) have recently gained attention for their lightweight nature, compactness, and fast decoding, showing promise for edge computing in consumer devices. Specifically, INR-based image compression methods implicitly store each image within a lightweight neural network, which serves as a compact representation of the image. However, most existing methods are limited to representing single-quality images with fixed-size models, which necessitates training separate models independently for images at varying quality levels, leading to additional training and storage costs. To tackle this problem, we propose a progressive image compression method based on Width-Depth Scalable Implicit Neural Representation (WDS-INR), which are composed of executable sub-networks of varying scales. By adjusting the scale of the sub-networks, WDS-INR can represent images at different quality levels while supporting progressive transmission. The scalable architecture of WDS-INR makes it well-suited for deployment on mobile and IoTs devices. Furthermore, we propose a band-limited initialization scheme that enhances both the representation capabilities and training stability of the WDS-INR. Finally, we introduce a meta-learning approach to the base sub-network to accelerate encoding $(4 times text { faster})$ . Experimental results demonstrate that the proposed method outperforms the baseline in rate-distortion performance $(+ 0.28~dB {~text {PSNR}})$ , while enabling scalable bit-rates with progressive decoding.
使用隐式神经表示（INR）的图像压缩方法最近因其轻量级、紧凑性和快速解码而受到关注，在消费设备的边缘计算中显示出前景。具体来说，基于inr的图像压缩方法隐式地将每个图像存储在一个轻量级的神经网络中，该神经网络作为图像的紧凑表示。然而，大多数现有方法仅限于用固定大小的模型表示单一质量的图像，这需要为不同质量水平的图像独立训练单独的模型，从而导致额外的训练和存储成本。为了解决这个问题，我们提出了一种基于宽度-深度可扩展隐式神经表示（WDS-INR）的渐进式图像压缩方法，该方法由不同规模的可执行子网络组成。通过调整子网的规模，WDS-INR可以表示不同质量水平的图像，同时支持逐行传输。WDS-INR的可扩展架构使其非常适合在移动和物联网设备上部署。此外，我们提出了一种带限初始化方案，增强了WDS-INR的表示能力和训练稳定性。最后，我们在基本子网络中引入了一种元学习方法来加速编码$(4 times text {faster})$。实验结果表明，该方法在率失真性能$(+ 0.28~dB {~text {PSNR}})$上优于基线，同时实现了可扩展的比特率和渐进解码。
{"title":"Lightweight Width-Depth Scalable Implicit Neural Representation for Progressive Image Compression","authors":"Qingyu Mao;Wenming Wang;Yongsheng Liang;Chenhu Xiao;Fanyang Meng;Gwanggil Jeon","doi":"10.1109/TCE.2025.3565495","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565495","url":null,"abstract":"Image compression approaches using implicit neural representation (INR) have recently gained attention for their lightweight nature, compactness, and fast decoding, showing promise for edge computing in consumer devices. Specifically, INR-based image compression methods implicitly store each image within a lightweight neural network, which serves as a compact representation of the image. However, most existing methods are limited to representing single-quality images with fixed-size models, which necessitates training separate models independently for images at varying quality levels, leading to additional training and storage costs. To tackle this problem, we propose a progressive image compression method based on Width-Depth Scalable Implicit Neural Representation (WDS-INR), which are composed of executable sub-networks of varying scales. By adjusting the scale of the sub-networks, WDS-INR can represent images at different quality levels while supporting progressive transmission. The scalable architecture of WDS-INR makes it well-suited for deployment on mobile and IoTs devices. Furthermore, we propose a band-limited initialization scheme that enhances both the representation capabilities and training stability of the WDS-INR. Finally, we introduce a meta-learning approach to the base sub-network to accelerate encoding <inline-formula> <tex-math>$(4 times text { faster})$ </tex-math></inline-formula>. Experimental results demonstrate that the proposed method outperforms the baseline in rate-distortion performance <inline-formula> <tex-math>$(+ 0.28~dB {~text {PSNR}})$ </tex-math></inline-formula>, while enabling scalable bit-rates with progressive decoding.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"5666-5675"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144867991","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Stable consumer electronic systems can assist traffic better. Good traffic consumer electronic systems require collaborative work between traffic algorithms and hardware. However, performance of popular traffic algorithms containing vehicle detection methods based on deep networks via learning data relation rather than learning differences in different lighting and occlusions is limited. In this paper, we present a dynamic Transformer network for vehicle detection (DTNet). DTNet utilizes a dynamic convolution to guide a deep network to dynamically generate weights to enhance adaptability of an obtained detector. Taking into relations of different information account, a mixed attention mechanism based channel attention and Transformer is exploited to strengthen relations of channels and pixels to extract more salient information for vehicle detection. To overcome the drawback of difference in an image account, a translation variant convolution relies on spatial location information to refine obtained structural information for vehicle detection. Experimental results illustrate that our DTNet is competitive for vehicle detection. Code of the proposed DTNet can be obtained at https://github.com/hellloxiaotian/DTNet.
{"title":"A Dynamic Transformer Network for Vehicle Detection","authors":"Chunwei Tian;Kai Liu;Bob Zhang;Zhixiang Huang;Chia-Wen Lin;David Zhang","doi":"10.1109/TCE.2025.3565318","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565318","url":null,"abstract":"Stable consumer electronic systems can assist traffic better. Good traffic consumer electronic systems require collaborative work between traffic algorithms and hardware. However, performance of popular traffic algorithms containing vehicle detection methods based on deep networks via learning data relation rather than learning differences in different lighting and occlusions is limited. In this paper, we present a dynamic Transformer network for vehicle detection (DTNet). DTNet utilizes a dynamic convolution to guide a deep network to dynamically generate weights to enhance adaptability of an obtained detector. Taking into relations of different information account, a mixed attention mechanism based channel attention and Transformer is exploited to strengthen relations of channels and pixels to extract more salient information for vehicle detection. To overcome the drawback of difference in an image account, a translation variant convolution relies on spatial location information to refine obtained structural information for vehicle detection. Experimental results illustrate that our DTNet is competitive for vehicle detection. Code of the proposed DTNet can be obtained at <uri>https://github.com/hellloxiaotian/DTNet</uri>.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"2387-2394"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-04-29DOI: 10.1109/TCE.2025.3565287
Tong Su;Yanan Jiang;Cuihua Hu
This paper presents a hybrid approach for energy management in smart camera systems by combining Convolutional Neural Networks (CNNs) for depth estimation, Stochastic Gradient Descent (SGD) for training, and modified Particle Swarm Optimization (PSO) for minimizing energy consumption. The proposed model deploys CNNs to accurately predict depth maps, while SGD fine-tunes the network’s weights to enhance prediction accuracy. On the other hand, the modified PSO algorithm optimizes camera settings to achieve energy savings without sacrificing depth estimation performance. The modification method consists of crossover and mutation operators to boost the global search ability of the algorithm. Experimental results on benchmark datasets, including KITTI and NYU Depth V2, demonstrate that the proposed hybrid model could successfully make a high reduction in energy consumption, with minimal loss in depth accuracy. Comparisons with baseline models show significant improvements in both energy efficiency and depth estimation precision.
{"title":"Machine Learning-Based Depth Estimation for Energy Optimization in Smart Camera Systems","authors":"Tong Su;Yanan Jiang;Cuihua Hu","doi":"10.1109/TCE.2025.3565287","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565287","url":null,"abstract":"This paper presents a hybrid approach for energy management in smart camera systems by combining Convolutional Neural Networks (CNNs) for depth estimation, Stochastic Gradient Descent (SGD) for training, and modified Particle Swarm Optimization (PSO) for minimizing energy consumption. The proposed model deploys CNNs to accurately predict depth maps, while SGD fine-tunes the network’s weights to enhance prediction accuracy. On the other hand, the modified PSO algorithm optimizes camera settings to achieve energy savings without sacrificing depth estimation performance. The modification method consists of crossover and mutation operators to boost the global search ability of the algorithm. Experimental results on benchmark datasets, including KITTI and NYU Depth V2, demonstrate that the proposed hybrid model could successfully make a high reduction in energy consumption, with minimal loss in depth accuracy. Comparisons with baseline models show significant improvements in both energy efficiency and depth estimation precision.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"4751-4758"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144867666","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-04-29DOI: 10.1109/TCE.2025.3565573
Yong Han;Xiaoliang Zhang;Jie Liu;Guangchun Liu;Weitao Yan
This paper develops a novel hybrid model based on Generative Adversarial Networks (GANs) and Differential Evolution (DE) to enhance remote sensing data and optimize resource assessment models for renewable energy management. GANs were employed to improve the resolution and quality of satellite imagery, addressing the challenges of low-resolution data and incomplete information. Quantitative evaluations, including Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM), demonstrated significant improvements in image quality, facilitating more accurate site assessments and predictive modeling. DE was applied to optimize key parameters such as sensor configurations and image enhancement algorithms, leading to enhanced accuracy in resource maps and reduced operational costs. The hybridization of GANs and DE created a comprehensive workflow that allowed for improved decision-making and efficient deployment. The proposed hybrid framework was shown to achieve higher prediction accuracy, exemplified by performance metrics such as Mean Absolute Error and R-squared values. Simulation results on case studies highlighted successful applications in renewable energy projects, emphasizing the potential of this integrated approach to drive cost-effective and scalable solutions.
{"title":"Application of Remote Sensing Technologies in Monitoring and Managing Renewable Energy Sources","authors":"Yong Han;Xiaoliang Zhang;Jie Liu;Guangchun Liu;Weitao Yan","doi":"10.1109/TCE.2025.3565573","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565573","url":null,"abstract":"This paper develops a novel hybrid model based on Generative Adversarial Networks (GANs) and Differential Evolution (DE) to enhance remote sensing data and optimize resource assessment models for renewable energy management. GANs were employed to improve the resolution and quality of satellite imagery, addressing the challenges of low-resolution data and incomplete information. Quantitative evaluations, including Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM), demonstrated significant improvements in image quality, facilitating more accurate site assessments and predictive modeling. DE was applied to optimize key parameters such as sensor configurations and image enhancement algorithms, leading to enhanced accuracy in resource maps and reduced operational costs. The hybridization of GANs and DE created a comprehensive workflow that allowed for improved decision-making and efficient deployment. The proposed hybrid framework was shown to achieve higher prediction accuracy, exemplified by performance metrics such as Mean Absolute Error and R-squared values. Simulation results on case studies highlighted successful applications in renewable energy projects, emphasizing the potential of this integrated approach to drive cost-effective and scalable solutions.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"4729-4735"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868334","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study focuses on optimal energy management of smart pocket parks in smart city for promoting elderly vitality based on consumer electronic devices. Recognizing the challenges elderly individuals face in maintaining an active lifestyle, the proposed system integrates innovative smart energy management system to provide energy demand of personalized sports activities compatible with users’ needs. In the proposed Deep Q-Learning (DQL) smart model, the system continuously learns from user interactions and health data to tailor recommendations, ensuring optimal energy consumption of devices and improvement in fitness levels. A key feature of this approach is the emphasis on energy management optimization for the devices used in the park. Through smart energy solutions, including solar-powered devices and efficient energy usage models, the system minimizes the environmental impact while ensuring consistent performance. Simulated Annealing (SA) is employed to fine-tune system parameters and avoid energy wastage, ensuring that devices operate at peak efficiency. The combination of DQL and SA allows for a synergistic system that not only promotes physical activity but also ensures sustainability and energy efficiency in smart cities. This methodology sets the stage for future advancements in urban health solutions for the elderly.
{"title":"An Intelligent Framework for Optimal Consumer Electronics’ Management in Smart Pocket Parks for Stimulating the Vitality of the Elderly","authors":"Lingling Zhu;Noor Aimran Samsudin;Zuhra Junaida Mohamad Husny Hamid;Haipeng Xu","doi":"10.1109/TCE.2025.3565411","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565411","url":null,"abstract":"This study focuses on optimal energy management of smart pocket parks in smart city for promoting elderly vitality based on consumer electronic devices. Recognizing the challenges elderly individuals face in maintaining an active lifestyle, the proposed system integrates innovative smart energy management system to provide energy demand of personalized sports activities compatible with users’ needs. In the proposed Deep Q-Learning (DQL) smart model, the system continuously learns from user interactions and health data to tailor recommendations, ensuring optimal energy consumption of devices and improvement in fitness levels. A key feature of this approach is the emphasis on energy management optimization for the devices used in the park. Through smart energy solutions, including solar-powered devices and efficient energy usage models, the system minimizes the environmental impact while ensuring consistent performance. Simulated Annealing (SA) is employed to fine-tune system parameters and avoid energy wastage, ensuring that devices operate at peak efficiency. The combination of DQL and SA allows for a synergistic system that not only promotes physical activity but also ensures sustainability and energy efficiency in smart cities. This methodology sets the stage for future advancements in urban health solutions for the elderly.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"4670-4676"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-04-29DOI: 10.1109/TCE.2025.3565308
Jingnan Duan;Jun Li
This research investigates the integration of Deep Q-Networks (DQN) and Ant Colony Optimization (ACO) in developing a hybrid framework aimed at optimizing energy consumption in computer vision systems, while ensuring performance in image processing tasks. With the increasing computational demands of modern applications, energy efficiency has become critical. The proposed methodology combines DQN’s ability to enhance decision-making processes with ACO’s exploration and exploitation strategies, resulting in efficient energy management. Key findings indicate that the DQN-ACO approach reduces energy consumption by 30% compared to baseline methods, while simultaneously improving accuracy by 8.24% and processing speed by 20%. Moreover, the framework exhibits strong adaptability in dynamic environments, leading to improvements in throughput and overall system performance. The outcomes of this research have significant implications for industries such as autonomous driving, surveillance, and mobile computing, where energy efficiency is paramount. By advancing both artificial intelligence and computer vision, this study contributes to the development of sustainable technology solutions, laying the foundation for future innovations in energy-efficient intelligent systems.
{"title":"Artificial Intelligence-Enabled Image Processing for Energy Optimization in Computer Vision Systems","authors":"Jingnan Duan;Jun Li","doi":"10.1109/TCE.2025.3565308","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565308","url":null,"abstract":"This research investigates the integration of Deep Q-Networks (DQN) and Ant Colony Optimization (ACO) in developing a hybrid framework aimed at optimizing energy consumption in computer vision systems, while ensuring performance in image processing tasks. With the increasing computational demands of modern applications, energy efficiency has become critical. The proposed methodology combines DQN’s ability to enhance decision-making processes with ACO’s exploration and exploitation strategies, resulting in efficient energy management. Key findings indicate that the DQN-ACO approach reduces energy consumption by 30% compared to baseline methods, while simultaneously improving accuracy by 8.24% and processing speed by 20%. Moreover, the framework exhibits strong adaptability in dynamic environments, leading to improvements in throughput and overall system performance. The outcomes of this research have significant implications for industries such as autonomous driving, surveillance, and mobile computing, where energy efficiency is paramount. By advancing both artificial intelligence and computer vision, this study contributes to the development of sustainable technology solutions, laying the foundation for future innovations in energy-efficient intelligent systems.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"4684-4691"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868228","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-04-29DOI: 10.1109/TCE.2025.3565680
Mingchuan Tan;Rencan Nie;Jinde Cao;Ying Zhang
Infrared and visible image fusion (IVIF) aims to merge images from both modalities of the same scene into a single image, enabling comprehensive information display and better support for visual computing tasks. Nevertheless, existing methods often overlook pixel-level relationships and struggle to effectively eliminate redundant information. To this end, we propose SMDFusion, a novel framework for fusing infrared and visible images using cross-modal noise-masked encoding and cross-modal differential perception information coupling. The framework consists of a self-supervised learning network (SSLN) and an unsupervised fusion network (UFN). Regarding the SSLN, the noise random masked encoder learns pixel-level relationships by employing a grid structure for multi-scale feature mapping that facilitates information exchange among different scales. The network is optimized with a self-supervision strategy for better representation learning. As for the UFN, symmetrical grid structures and multi-scale attention mechanisms are utilized to integrate intra-modal features while the cross-modal difference perception (CDP) module eliminates redundant information between modalities and conditionally captures complementary perception. The fusion image is synthesized by computing the modality-specific contribution estimation. Qualitative and quantitative experimental results demonstrate that SMDFusion outperforms representative methods in the task of multi-modal information fusion as well as supporting downstream tasks. The code is available at:https://github.com/rcnie/IVIF-SMDFusion.
{"title":"SMDFusion: A Self-Supervised Fusion for Infrared and Visible Images via Cross-Modal Random Noise Masked Encoding and Difference Perception","authors":"Mingchuan Tan;Rencan Nie;Jinde Cao;Ying Zhang","doi":"10.1109/TCE.2025.3565680","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565680","url":null,"abstract":"Infrared and visible image fusion (IVIF) aims to merge images from both modalities of the same scene into a single image, enabling comprehensive information display and better support for visual computing tasks. Nevertheless, existing methods often overlook pixel-level relationships and struggle to effectively eliminate redundant information. To this end, we propose SMDFusion, a novel framework for fusing infrared and visible images using cross-modal noise-masked encoding and cross-modal differential perception information coupling. The framework consists of a self-supervised learning network (SSLN) and an unsupervised fusion network (UFN). Regarding the SSLN, the noise random masked encoder learns pixel-level relationships by employing a grid structure for multi-scale feature mapping that facilitates information exchange among different scales. The network is optimized with a self-supervision strategy for better representation learning. As for the UFN, symmetrical grid structures and multi-scale attention mechanisms are utilized to integrate intra-modal features while the cross-modal difference perception (CDP) module eliminates redundant information between modalities and conditionally captures complementary perception. The fusion image is synthesized by computing the modality-specific contribution estimation. Qualitative and quantitative experimental results demonstrate that SMDFusion outperforms representative methods in the task of multi-modal information fusion as well as supporting downstream tasks. The code is available at:<uri>https://github.com/rcnie/IVIF-SMDFusion</uri>.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"2579-2591"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868391","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-04-29DOI: 10.1109/TCE.2025.3565590
Yihong Li;Qiang Song
This article explores the integration of Artificial Intelligence (AI) and Big Data Analytics to optimize energy consumption in IoT-enabled smart home devices. It presents a robust analytical framework that leverages Variational Autoencoders (VAEs) for feature extraction and Differential Evolution (DE) for optimizing energy management parameters. Data was gathered from various IoT devices, including energy usage patterns, occupancy data, and environmental conditions. The results show a notable 40% reduction in energy consumption, leading to annual cost savings of up to ${$}300$ per household. Moreover, user satisfaction increased by 25%, with participants reporting heightened awareness and engagement in energy conservation. The study highlights how the proposed framework efficiently identifies common usage patterns and optimizes energy distribution while preserving user comfort. These findings reinforce the potential of AI-driven analytics in improving energy efficiency in smart homes, demonstrating that advanced algorithms not only support energy conservation but also promote active user participation in sustainability efforts.
{"title":"Big Data-Intelligence Analytics for Energy Optimization in IoT-Enabled Smart Home Devices","authors":"Yihong Li;Qiang Song","doi":"10.1109/TCE.2025.3565590","DOIUrl":"https://doi.org/10.1109/TCE.2025.3565590","url":null,"abstract":"This article explores the integration of Artificial Intelligence (AI) and Big Data Analytics to optimize energy consumption in IoT-enabled smart home devices. It presents a robust analytical framework that leverages Variational Autoencoders (VAEs) for feature extraction and Differential Evolution (DE) for optimizing energy management parameters. Data was gathered from various IoT devices, including energy usage patterns, occupancy data, and environmental conditions. The results show a notable 40% reduction in energy consumption, leading to annual cost savings of up to <inline-formula> <tex-math>${$}300$ </tex-math></inline-formula> per household. Moreover, user satisfaction increased by 25%, with participants reporting heightened awareness and engagement in energy conservation. The study highlights how the proposed framework efficiently identifies common usage patterns and optimizes energy distribution while preserving user comfort. These findings reinforce the potential of AI-driven analytics in improving energy efficiency in smart homes, demonstrating that advanced algorithms not only support energy conservation but also promote active user participation in sustainability efforts.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"4721-4728"},"PeriodicalIF":10.9,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868016","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The proliferation of Internet of Things (IoT) and embedded computing has led to widespread deployment of smart consumer electronics requiring edge-based Artificial Intelligence (AI) capabilities. However, the heterogeneous nature of sensing data and dynamic edge environments poses significant challenges for efficient model inference on resource-constrained devices. To address these challenges, this paper presents a lightweight collaborative inference framework designed for consumer electronics. First, we formulate the inference optimization problem as a Mixed-Integer Nonlinear Programming (MINLP) problem, considering channel pruning, early exit and cloud offloading decisions to optimize the trade-off between accuracy and computational cost. Second, we propose a selective model activation mechanism based on Markov Decision Process (MDP), which employs a recursive self-attention mechanism to dynamically track inference budgets and guide decision-making through encoder-decoder architectures. The mechanism integrates entropy regularization during training to ensure robust and diverse execution paths. Comprehensive experiments demonstrate that our framework achieves 65.50% reduction in model parameters and 80.68% reduction in inference Floating Point Operations (FLOPs) while maintaining accuracy loss within 0.81% of the original model, making it suitable for real-time AI applications on resource-constrained consumer electronics.
{"title":"A Lightweight Cloud-Edge Collaborative Intelligence Inference Framework With Runtime Dynamic Optimization for Resource-Constrained Consumer Electronics","authors":"Chenlu Wang;Yuhuai Peng;Dawei Zhang;Ryan Alturki;Bandar Alshawi;Majid Alotaibi","doi":"10.1109/TCE.2025.3564777","DOIUrl":"https://doi.org/10.1109/TCE.2025.3564777","url":null,"abstract":"The proliferation of Internet of Things (IoT) and embedded computing has led to widespread deployment of smart consumer electronics requiring edge-based Artificial Intelligence (AI) capabilities. However, the heterogeneous nature of sensing data and dynamic edge environments poses significant challenges for efficient model inference on resource-constrained devices. To address these challenges, this paper presents a lightweight collaborative inference framework designed for consumer electronics. First, we formulate the inference optimization problem as a Mixed-Integer Nonlinear Programming (MINLP) problem, considering channel pruning, early exit and cloud offloading decisions to optimize the trade-off between accuracy and computational cost. Second, we propose a selective model activation mechanism based on Markov Decision Process (MDP), which employs a recursive self-attention mechanism to dynamically track inference budgets and guide decision-making through encoder-decoder architectures. The mechanism integrates entropy regularization during training to ensure robust and diverse execution paths. Comprehensive experiments demonstrate that our framework achieves 65.50% reduction in model parameters and 80.68% reduction in inference Floating Point Operations (FLOPs) while maintaining accuracy loss within 0.81% of the original model, making it suitable for real-time AI applications on resource-constrained consumer electronics.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"6041-6054"},"PeriodicalIF":10.9,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-04-28DOI: 10.1109/TCE.2025.3563698
Huazhong Liu;Weiyuan Zhang;Ren Li;Yunfan Zhang;Jihong Ding;Guangshun Zhang;Hanning Zhang;Laurence T. Yang
With the continuous expansion of consumer electronics applications, various data generated from ubiquitous consumer electronics devices are experiencing exponential growth. By leveraging the significant advantages of multidimensional association analysis, tensor-based big data technology has proven effective in uncovering hidden patterns within these data. However, the curse of dimensionality severely restricts the widespread exploitation of tensors, particularly on edge devices with limited computing and storage capabilities under cloud-edge computing environments. To address this challenge, we propose a series of cloud-edge collaborative scalable Tucker-based tensor computations to effectively analyze these ubiquitous data. First, we present a set of Tucker-based tensor operations that transform high-order and large-scale tensor operations into multiple low-order and small-scale operations while preserving the equivalence of their results. Then, we present a scalable Tucker-based computation architecture to adapt to the cloud-edge computing paradigm, including scalable inter-Tuckercore and intra-Tuckercore models. Furthermore, we implement some typical Tucker-based tensor computations based on various scalable models and analyze their complexity in detail. Finally, extensive evaluations on both synthetic and real-world datasets demonstrate that the proposed scalable Tucker-based tensor computation method significantly improves computational efficiency, achieving an average efficiency improvement of 2 to 5 times compared to serial computation. These results confirm its suitability for cloud-edge collaboration to process ubiquitous consumer electronics data.
{"title":"Cloud-Edge Collaborative Scalable Tucker-Based Tensor Computations for Ubiquitous Consumer Electronics Data","authors":"Huazhong Liu;Weiyuan Zhang;Ren Li;Yunfan Zhang;Jihong Ding;Guangshun Zhang;Hanning Zhang;Laurence T. Yang","doi":"10.1109/TCE.2025.3563698","DOIUrl":"https://doi.org/10.1109/TCE.2025.3563698","url":null,"abstract":"With the continuous expansion of consumer electronics applications, various data generated from ubiquitous consumer electronics devices are experiencing exponential growth. By leveraging the significant advantages of multidimensional association analysis, tensor-based big data technology has proven effective in uncovering hidden patterns within these data. However, the curse of dimensionality severely restricts the widespread exploitation of tensors, particularly on edge devices with limited computing and storage capabilities under cloud-edge computing environments. To address this challenge, we propose a series of cloud-edge collaborative scalable Tucker-based tensor computations to effectively analyze these ubiquitous data. First, we present a set of Tucker-based tensor operations that transform high-order and large-scale tensor operations into multiple low-order and small-scale operations while preserving the equivalence of their results. Then, we present a scalable Tucker-based computation architecture to adapt to the cloud-edge computing paradigm, including scalable inter-Tuckercore and intra-Tuckercore models. Furthermore, we implement some typical Tucker-based tensor computations based on various scalable models and analyze their complexity in detail. Finally, extensive evaluations on both synthetic and real-world datasets demonstrate that the proposed scalable Tucker-based tensor computation method significantly improves computational efficiency, achieving an average efficiency improvement of 2 to 5 times compared to serial computation. These results confirm its suitability for cloud-edge collaboration to process ubiquitous consumer electronics data.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 2","pages":"4026-4038"},"PeriodicalIF":10.9,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144868183","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：