Honghao Gao’s research while affiliated with Shanghai University and other places

With consumer electronics development, consumerelectronic WiFi-based human pose estimation (HPE) has been acknowledged as an emerging and privacy-friendly technology. Since WiFi signals do not directly capture the human image, WiFi offers a suitable solution for various scenarios. Such as security monitoring and privacy-preserving human monitoring. Existing studies have neglected the sparsity in the joint heatmaps of HPE, and also have not effectively utilized the diversity ofWiFi signals. As a result, it is difficult to provide an accurate human pose estimation. To overcome the above challenges, we propose a sparse regularization method for WiFi-based HPE (WiPE). Considering the sparse nature of the joint heatmap, we design a sparse regularized neural network to focus output around the joint. Subsequently, through a three-dimensional streaming signal fusion modular, we sufficiently integrate diverse features in subcarriers ofWiFi signals, obtaining key frames. Experiment results on a real establishing system show that our WiPE outperforms state-of-the-art methods. WiPE achieved a PCK@0.2 score of 95.97%, indicating high accuracy. At the same time, our WiPE is more lightweight than other methods. WiPE has 43.14 floating point operations (FLOPs). Morphological information of the body is also obtained by WiPE.

In-band network telemetry (INT) is a new network measurement technique that provides real-time, fine-grained packet-level network measurements. However, standard INT lacks the flexibility to perform configurable on-demand network measurements. In this work, we propose a flexible on-demand network measurement mechanism (FDSR-INT) based on a dual bitmap by combining the programmability of segment routing based on IPv6 (SRv6) and the telemetry efficiency of the INT to achieve customizable network measurements. By designing the dual bitmap, the telemetry information of personalized probe nodes is supported, and the telemetry efficiency is improved. SRv6 is employed to direct measurement probe packets, ensuring coverage of a specified set of network targets. Its inherent programmability enables INT to conduct customized per-hop network measurements. The designed flexible INT field structure, which appends traceability information along with telemetry information, reduces the information traceability overhead in the control plane and improves the network compatibility of the mechanism. We solve the optimal probing path by solving a traveler problem in an auxiliary graph and design a greedy pathcutting algorithm to maximize the number of nodes for single packet probing while satisfying the packet length constraints to improve the success rate of the probing task. Finally, we implemented FDSR-INT using P4 and verified its performance experimentally in the constructed space.air.ground integrated simulation dynamic environment. FDSR-INT saves 30% of the data plane bandwidth and 54% of the data plane bandwidth on north.south interfaces compared with SONM-SR-INT, etc. Furthermore, it has low control plane processing overhead and probe path transmission overhead.

Brain tumors require AI-assisted, precise treatments. Methods based on healthcare electronics, such as magnetic resonance imaging (MRI), computed tomography (CT), and gastrointestinal endoscopy, are widely used in hospitals. Maximizing the functionality of these electronic devices is crucial for improving tumor lesion detection and is a challenge in AI-assisted medical applications. Precise classification is vital for effectively planning brain tumor treatments, and accurate classification results offer crucial insights that enable physicians to devise optimal treatment strategies. Therefore, this paper proposes a novel brain tumor classification method named SwinBTC, which is based on the healthcare Internet of Things (HIoT) and integrates a pretrained and fine-tuned Swin transformer model to improve the performance of healthcare electronics. First, a transfer learning (TL)-based brain tumor classification model called SwinBTC is proposed. The general features in the images are used to improve the classification ability of the brain tumor MRI results, further accelerate the model training speed, and avoid overfitting problems. Second, clinical magnetic resonance images obtained from HIoT nodes are used to enrich the dataset, and online and offline data augmentation techniques are used to expand the utilized dataset and increase data diversity, improving the generalizability and reliability of the developed model. Finally, the performance of the proposed approach is evaluated on the CE-MRI and TT-MRI datasets via various classification metrics, and the experimental results reveal that our method outperforms other baselines.

As communication and computing technologies advance, vehicular edge computing emerges as a promising paradigm for delivering a wide array of intelligent services in 6G enabled Intelligent Autonomous Transport Systems. These service requests, are safety-oriented and typically require the fusion of processing results from multiple independent computation tasks generated by various onboard sensors, in which the computation tasks are delay-sensitive and computation-intensive. Consequently, the allocation of multiple tasks within a single service request while efficiently reducing request completion time and energy consumption presents a substantial challenge. In order to address the problem of multi-task simultaneous scheduling, this paper proposed to employ deep reinforcement learning and edge computing architecture to make task scheduling decisions for vehicles. Firstly, the Vehicle-Infrastructure Network (VINET) is designed, in which the vehicles can assign multiple tasks to the edge servers and other idle vehicles, thus extending the task processing capabilities for vehicles. Secondly, Fully-decentralized Multi-agent Proximal Policy Optimization (FMPPO) algorithm is proposed to make task scheduling decisions for autonomous driving, the large model trained via FMPPO is adaptable to different scenarios with various numbers of vehicles. Thirdly, by taking into account task characteristic, environmental status, and vehicle mobility, the proposed method can make task scheduling decisions in real-time and then dynamically distributes tasks based on the decisions. Finally, experimental results demonstrate that the designed method outperforms benchmark methods in terms of both completion time and energy consumption of computation tasks.

Deploying Unmanned Aerial Vehicles (UAVs) as aerial base stations enhances the coverage and performance of communication networks in Vehicular Edge Computing (VEC) scenarios. However, due to the limited communication range and energy capacity of UAVs, they cannot continuously cover entire areas or sustain long flights. Therefore, achieving full communication coverage of a target area with a minimal number of UAVs and efficient task offloading remains a significant challenge. To address this problem, the UAV-assisted Two-stage Intelligent Collaboration (UTIC) method is proposed in this paper to tackle the joint position optimization and task scheduling issue. Firstly, a UAV-assisted Two-stage Task Scheduling (UTTS) system model is designed to optimize the allocation process. Secondly, an Enhanced Particle Swarm Optimization (E-PSO) algorithm is designed to determine the optimal positions of UAVs, ensuring complete coverage of all mobile vehicles (MVs) with the minimum number of UAVs. Thirdly, Deep Deterministic Policy Gradient (DDPG) method is employed to find the optimal scheduling decisions for MVs, considering energy consumption, delay, and task priorities. Simulation results demonstrate that the proposed UTIC method can achieve nearly 20% reduction in UAV deployment and outperform three other classical reinforcement learning (RL) algorithms in terms of reducing system cost.

Deep learning models are widely used to process Computed Tomography (CT) data in the automated screening of pulmonary diseases, significantly reducing the workload of physicians. However, the three-dimensional nature of CT volumes involves an excessive number of voxels, which significantly increases the complexity of model processing. Previous screening approaches often overlook this issue, which undoubtedly reduces screening efficiency. Towards efficient and effective screening, we design a hierarchical approach to reduce the computational cost of pulmonary disease screening. The new approach re-organizes the screening workflows into three steps. First, we propose a Computed Tomography Volume Compression (CTVC) method to select a small slice subset that comprehensively represents the whole CT volume. Second, the selected CT slices are used to detect pulmonary diseases coarsely via a lightweight classification model. Third, an uncertainty measurement strategy is applied to identify samples with low diagnostic confidence, which are re-detected by radiologists. Experiments on two public pulmonary disease datasets demonstrate that our approach achieves comparable accuracy and recall while reducing the time by 50%-70% compared with the counterparts using full CT volumes. Besides, we also found that our approach outperforms previous cutting-edge CTVC methods in retaining important indications after compression.

In intelligent transportation systems, data routing in vehicle-to-everything (V2X) networks is key to ensuring efficient information transfer among vehicles, pedestrians and infrastructure. The quality of data routing directly affects communication efficiency and system performance. However, data routing in V2X networks often faces potential security threats, which may lead to communication interruption, data delay or information loss. Unreliable routing fails to meet the communication quality of service (QoS) requirements for V2X networks. Therefore, this paper proposes a joint scheme that combines trust-prediction and attack-resistance to ensure reliable routing in V2X networks. First, this scheme employs a fuzzy control-based trust evaluation method to provide direct trust indicators. Second, a trust prediction method based on deep belief networks is utilized to evaluate vehicle status. A classification scheme based on trust levels is used to filter candidate sets for network repair to help the network resist malicious behaviour. Last, a novel routing decision function is introduced to plan reliable routes. Routes planned on the basis of this function not only meet the basic requirements of reliable routing, but are also suitable for routing requirements in different scenarios, such as minimizing transmission latency. The Experimental results show that, compared with the three baseline schemes, this scheme improves the accuracy and false alarm rate on the UNSW-NB15 dataset by 2.94% and 6.31%, respectively, and this scheme also performs better in terms of the data reception rate and transmission delay rate in actual application scenarios.

Monocular 3D object detection is essential for identifying objects in road images, thus offering valuable environmental perception data that are crucial for human-centric autonomous driving systems. However, due to the inherent limitations of camera imaging, obtaining precise depth information from images alone is challenging, which hampers the accuracy of within-scene object localization. In this paper, we introduce a monocular 3D object detection method called MonoFG that uses knowledge distillation with separated foreground and background components to improve the accuracy of object localization. First, detached foreground and background distillation processes can strategically leverage the distinct positional information acquired from each location to optimize the produced global distillation effects. This step serves as the foundation for the subsequent feature and response distillation process, which focuses on the distilled foreground and background rather than isolated object distillation. Second, triple attention mechanism-based feature distillation intensifies the feature imitation and feature representation capabilities of the student network. Spatial and channel attention mechanisms encourage the student network to capture crucial pixels and channels from the teacher network, whereas a self-attention mechanism globally transfers the learned relationships between pixels. Third, localization error-based response distillation facilitates a clearer transfer of positional information from the teacher network to the student network. Only when the positioning ability of the teacher network exceeds that of the student network can knowledge be comprehensively distilled across both the foreground and background. Therefore, the distillation process is constrained to specific content, which is delineated by positioning errors that serve as the boundaries. Finally, experiments conducted on the KITTI benchmark dataset demonstrate that our method outperforms many well-known baseline methods in several representative evaluation tasks (e.g., 3D object detection and bird's-eye view (BEV) detection) involving human-centric autonomous driving systems.