查看更多>>摘要：Non-intrusive load monitoring (NILM), a sophisticated load monitoring technology, has garnered considerable interest for its potential to assist consumers in lowering their energy expenditures. In this paper, we present a continuous non-convex optimization model for NILM that employs the norm-box constraint to convert the discrete integer variables in the model into continuous ones. Subsequently, we apply the alternating direction method of multipliers (ADMM) algorithm to tackle the non-convex problem. To enhance the sluggish convergence of the ADMM algorithm, we introduce a linear penalty term based on relaxation distance (RD) to supplant the conventional quadratic penalty term. Furthermore, we devise a heuristic refinement method based on neighborhood search (NS) to augment the solution quality of our algorithm. Simultaneously, by utilizing a dynamic window partitioning technique, the NILM task can be split into multiple small subtasks. These subtasks can be allocated to multiple consumer electronics with computing capabilities to achieve distributed computing. Ultimately, we validate our proposed algorithm on the AMPds dataset, and the experimental results demonstrate that it has faster convergence and yields better solutions compared to a state-of-the-art solver and traditional ADMM algorithms. Using our algorithm, the NILM system can offer consumers efficient, convenient, and economical services.
查看更多>>摘要：Thermal InfraRed (TIR) technology has achieved significant progress, in light of its ability to reflect lighting conditions in dark environments, enhancing its vital role in industrial and consumer electronics. However, current research on TIR image semantic segmentation mainly focuses on urban scenes, while the segmentation of human bodies in the TIR domain remains an under-explored area, which holds considerable promise for applications such as low-light security checks, nocturnal combat scenarios, and human action recognition. In this paper, we introduce a novel computer-vision task—Thermal InfraRed Human Body Parsing, which aims to generate accurate segmentation maps for different parts of human bodies in TIR images. To open up future research in this area, we collect a new dataset called HBTIR, which contains TIR images and corresponding semantic labels of 32 participants in various poses. Furthermore, we propose a novel neural network called HBTIR-Seg, which incorporates an edge-guided attention mechanism specifically tailored for TIR human body imagery. Extensive experiments demonstrate that our method greatly outperforms existing segmentation methods on the HBTIR dataset.
查看更多>>摘要：Image compression is essential for reducing the cost to save or transmit images. Recently, learned image compression methods have achieved superior compression performance compared to traditional image compression standards. Many learned image compression methods utilize convolutional entropy models to remove local spatial and channel redundancy in the latent representation. Some recent methods incorporate transformer to further eliminate non-local redundancy. However, these methods employ the same transformer structure to model both spatial and channel correlations, thereby failing to take advantage of the difference between the spatial characteristics and the channel characteristics of the latent representation. To resolve this issue, we propose novel adaptive channel and window-based spatial entropy models. The adaptive channel entropy model, which consists of the channel transformer module and the channel excitation module, dynamically fuses and excites channel information to implicitly predict channel context. More specifically, we first establish the relationship between the decoded channels and the channels to be encoded. Based on that channel relationship, the channel transformer module adaptively updates the predicted channel context. Finally, the channel excitation module is employed to emphasize informative channel context and suppress irrelevant channel context. Furthermore, we introduce a window-based spatial entropy model to capture global semantic information within the window and generate the spatial context of non-anchor features based on the decoded anchor features. The spatial context and channel context are combined to predict the Gaussian parameters of the latent representation. Experimental results demonstrate that our method outperforms some state-of-the-art image compression methods on Kodak, CLIC and Tecnick datasets.
查看更多>>摘要：The rapid development of e-Health provides elderly consumers with more convenient medical services. Alzheimer’s disease is one of the major diseases that threaten the health of the elderly. Its early detection is vital for its effective treatment and management. In this study, an end-to-end model, individual-to-group graph convolutional network (IG-GCN), is proposed for AD early detection and abnormal brain region identification. Specifically, the proposed IG-GCN first learns the low-dimensional brain graph embeddings of individual brain networks, and then incorporates individual non-imaging information to construct an information-rich group network for all participants. Experimental results demonstrate that the proposed model surpasses baseline methods in AD prediction and can effectively identify abnormal brain regions and biomarkers at various stages of the disease. The brain network-based IG-GCN framework not only advances pathological research and early treatment of AD, but is also more amenable to integration with consumer electronics due to its low dimensionality and simplicity compared to traditional brain imaging data, offering a novel avenue for smart healthcare solutions within the realm of consumer electronics.
查看更多>>摘要：Anomalies within data flows in the Internet of Things environment can potentially result in security vulnerabilities in consumer electronics. Therefore, it is crucial to effectively detect anomaly data to safeguard the reliability and continuous functionality of consumer electronics. Existing related works either learn from unlabeled data using unsupervised methods or leverage the limited labeled data to improve detection performance by semi-supervised methods. However, these methods usually overfit specific types of known anomalies or ignore the uncertainty when model training. To this end, we design a novel approach to jointly optimize the end-to-end detection of labeled and unlabeled anomalies. Specifically, the anomaly data detection problem investigated is first reformulated as a Markov decision process. Then, a partially labeled anomaly data detection approach (PANDA) based on prioritized deep deterministic policy gradient is proposed, which considers uncertainty when the agent makes decisions and can learn from the labeled known anomalies while continuously exploring and detecting prospective anomalies in unlabeled data. Extensive experiments on 13 datasets show that PANDA improves the AUC-ROC and AUC-PR by 3.0%-10.3% and 10.0%-73.5% and its robustness under the impact of anomaly contamination rates compared with four state-of-the-art competing methods.
查看更多>>摘要：Lithium-ion batteries (LIBs) are a widely used energy storage technology owing to their excellent energy density, minimal self-discharge property, and high cycle life. Despite these promising features, their performance is affected by both low and high temperatures. When the internal temperature exceeds a certain threshold, the battery may experience thermal runaway, leading to fire and explosion. Moreover, this process is accelerated at high charge/discharge currents. Therefore, in high current applications, accurate monitoring of the internal temperature of the battery becomes critically important to ensure the safety. Hence, an improved coupled electrothermal model (ICETM) has been proposed by combining a novel three-state thermal model with an existing electrical equivalent circuit model through temperature dependent electrical parameters and heat generation. The primary aim is to improve the accuracy of internal temperature estimation of the battery at high currents while accounting for time efficiency in thermal model parameterization. The ICETM is parameterized through experimental and simulation studies using a LiFePO4/graphite battery. The effectiveness of the proposed model and parameterization method is validated experimentally using two case studies. The results show 14% improvement in accuracy and 140–160 hours time reduction over its existing counterparts in estimating core temperature and model parameterization, respectively.
NETL
NSTL
IEEE