2 June 2025 in IEEE Transactions on Consumer Electronics
Electrocardiogram (ECG) readings play a vital role in diagnosing cardiovascular diseases, including myocardial infarction (MI), a condition that severely damages heart tissue and can lead to fatal outcomes. Consumer electronic devices are used to collect ECG signals, which reveal crucial details about MI. Timely and precise diagnosis is essential to reduce mortality, and this can be enhanced using advanced deep-learning models like ResNet. This paper introduces the CardioBERT, designed to detect cardiovascular disease from ECG signals using Convolutional Neural Network (CNN) and large language models (LLMs) like BERT. Since CNN is traditionally built for multidimensional data, whereas ECG signals are inherently one-dimensional, our CardioBERT employs residue-level contact-map predictions to extract and optimally integrate features, effectively addressing the dimensionality mismatch. Furthermore, BERT enriches the feature fusion process by capturing and interpreting intricate patterns within the data. By employing consumer electronics and mathematical transformations (e.g., reciprocal and cubic functions), the CardioBERT achieves a notable 0.92% increase in accuracy with existing methods. This improvement underscores the potential of our CardioBERT, enhanced by LLMs, to advance cardiovascular healthcare systems significantly.
12 January 2026 in ACM Transactions on Multimedia Computing, Communications, and Applications
With the rapid development of smart device technology, the current version of the Internet of Things (IoT) is moving towards a multimedia IoT because of multimedia data. This innovative concept seamlessly integrates multimedia data with the IoT-Edge Continuum. Recently, a distributed learning framework has shown promise in revolutionizing various industries, including smart cities, healthcare, etc. However, these applications may face challenges, such as the presence of malicious devices that invade the privacy of other devices or corrupt uploaded model parameters. Additionally, the existing synchronous federated learning (FL) methods face challenges in effectively training models on local datasets due to the diversity of IoT devices. To tackle these concerns, we propose an efficient and privacy-enhanced asynchronous FL approach for multimedia data in edge-based IoT. In contrast to traditional FL methods, our approach combines revocable attribute-based encryption (RABE) and differential privacy (DP). This guarantees the privacy of the entire process while allowing seamless collaboration between multiple devices and the aggregation server during model training. Also, this combination brings a dynamic nature to the system. Furthermore, we utilize an asynchronous weight-based aggregation algorithm to improve the efficiency of training and the quality of the final returned model. Our proposed scheme is confirmed by theoretical safety proofs and experimental results with multimedia data. Performance evaluation shows that our framework reduces the cryptography runtime by 63.3% and the global model aggregation time by 61.9% compared to cutting-edge schemes. Moreover, our accuracy is comparable to the most primitive FL schemes, maintaining 86.7%, 70.8%, and 86.1% on MNIST, CIFAR-10, and Fashion-MNIST, respectively. The experimental results highlight the remarkable practicality, resilience and effectiveness of the proposed scheme.
16 October 2024 in IEEE Journal of Biomedical and Health Informatics
The application of computer vision-powered large language models (LLMs) for medical image diagnosis has significantly advanced healthcare systems. Recent progress in developing symmetrical architectures has greatly impacted various medical imaging tasks. While CNNs or RNNs have demonstrated excellent performance, these architectures often face notable limitations of substantial losses in detailed information, such as requiring to capture global semantic information effectively and relying heavily on deep encoders and aggressive downsampling. This paper introduces a novel LLM-based Hybrid-Transformer Network (HybridTransNet) designed to encode tokenized Big Data patches with the transformer mechanism, which elegantly embeds multimodal data of varying sizes as token sequence inputs of LLMS. Subsequently, the network performs both inter-scale and intra-scale self-attention, processing data features through a transformer-based symmetric architecture with a refining module, which facilitates accurately recovering both local and global context information. Additionally, the output is refined using a novel fuzzy selector. Compared to other existing methods on two distinct datasets, the experimental findings and formal assessment demonstrate that our LLM-based HybridTransNet provides superior performance for brain tumor diagnosis in healthcare informatics.
25 April 2025 in IEEE Internet of Things Journal
The way digital assets are exchanged is rapidly evolving with the emergence of the Metaverse. In particular, immersive and decentralized platforms are reshaping transaction dynamics by offering new levels of interactivity. However, with these innovations come substantial challenges, especially those related to ensuring secure, energy-efficient, and interoperable asset flows across domains. Motivated by these needs, this article present a transaction framework that is designed to balance scalability and sustainability. Specifically, we combine the relay chain-parachain architecture with chameleon hash signatures (CHSs) to construct a mechanism capable of handling cross-domain digital content exchange. The proposed design aims to support flexible asset handling while keeping computation and energy demands in check. Taking inspiration from green supply chain management (GSCM), the framework also reflects concerns about sustainable digital asset governance. To evaluate its practicality, we carried out experiments across various network conditions and transaction volumes. The evaluation results suggest that our framework maintains low latency and reasonable energy efficiency, all while ensuring secure and interoperable transactions in the Metaverse context. These findings point to the potential of the framework as a stable infrastructure for digital asset management at scale in future decentralized environments.
10 December 2025 in IEEE Transactions on Consumer Electronics
Due to the influence of non-line-of-sight (NLOS) signals transmitted by tunnels, flyover, and urban forest environments, the precise positioning of connected and automated vehicles (CAVs) faces a great challenge. This paper proposes a second-order cone programming (SOCP) estimator for vehicle cooperative localization based on Polarized Identity fusion of the relative distance (RD) and the angle of arrival (AOA) between the target vehicle and the base station. First, the positioning error between the target CAVs and the base station is studied in a three dimensional (3D) driving environment. Second, with assumption of unknown and known NLOS status information, the optimal estimation analysis of SOCP is carried out using the polarized identity method. Meanwhile, the complexity and accuracy of the proposed algorithm are evaluated under different driving scenarios, both with and without the NLOS status information between the target CAVs and the base station. The simulation results show that the localization performance of the proposed method outperforms the state-of-the-art in the simulations environment of the vehicular ad hoc networks (VANETs). Finally, the experiment results demonstrate the ability and effectiveness of the proposed algorithm to handle NLOS scenarios with known status information.
4 December 2024 in IEEE Transactions on Consumer Electronics
Vehicular consumer electronics, such as autonomous vehicles (AVs), need collecting large amounts of private user information, which face the risk of privacy leakage. To protect the privacy of consumers, researchers have proposed to apply federated learning (FL) to privacy-preserving vehicular consumer electronics, that is, leveraging FL for collaborative training of a decision model without exchanging the sensitive information generated by AVs. However, FL generally needs to exchange the gradients of client models periodically with the central server, which can be attacked by adversaries to infer user information. Thereby, it may still face the risk of privacy leakage. To solve that challenge, we put forward a novel FL framework, called FL with Meta-Layers Training (FL-MLT). Instead of exchanging the gradients of the client models, it exchanges, meta-layers with the central servers. Since meta-layers are only a slice of the client models, exchanging them helps protect privacy. On the other hand, they contain meta-knowledge to help the FL training process. In the experiments, we conduct extensive visual classification simulation to evaluate FL-MLT, and the experimental results demonstrate the superior performance of FL-MLT.