Search

In advanced manufacturing, ergonomic risk assess-ment and fatigue analysis ensure not only the health of human operators, but also the productivity and quality of the manu-facturing process. In this paper we investigate a vision-based method for automatic ergonomic and fatigue risk monitoring, based on a cost-efficient 3D camera system and AI-driven video-based approaches for 3D body posture analysis and repetition counting. Our laboratory trials showed that this method was able to track joint motions with an average accuracy of 3.5°, performed comparable to a human operator when assessing the ergonomic risk, and was able to track and focus on repetitive motions of the human operator. The proposed method supports data visualization, real-time ergonomic and fatigue analyses, and report generation, and has the potential to support the development of better manufacturing environments., 2022 IEEE 5th International Conference on Industrial Cyber-Physical Systems (ICPS), May 24-26, 2022, Coventry, United Kingdom (Online Conference)

Nowadays, blockchain as emerging computer technology is gradually applied to the Internet of Things (IoT) industry. However, high traffic load and network congestion have become the main obstacles to the integration of IoT and blockchain, thanks to the vast amount of IoT devices as well as the considerable information redundancy of blockchain. In IoT-blockchain, the number of IoT devices and the blockchain network scale are supposed to establish an appropriate balance. Therefore, building a precise traffic model between the blockchain and IoT devices is crucial for designing and optimizing an IoT-Blockchain network. This paper proposes an innovative traffic model of the IoT-blockchain system, which describes traffic of blockchain nodes under the different sizes of IoT input and network scale. We implement a lightweight Bitcoin-like blockchain based on PoW and connect with several IoT simulators to conduct various network traffic experiments. Our mathematical model and experimental results demonstrate the relationship between self-similarity, network traffic, blockchain scales, and the number of IoT inputs.

In order to study the influencing factors of the submarine optical cable in the laying process, this paper builds a mathematical model of the submarine optical cable laying process on the basis of comprehensive consideration of various loads, derives the kinematic equations of the submarine optical cable, and uses the Rung-Kutta algorithm to solve the calculation. The effects of parameters such as the speed of the cable-laying ship and the speed of cable paying-out, diameter, and Poisson’s ratio on the laying process are studied. The results show that: the speed of the cable-laying ship, the speed of cable paying-out and the diameter have a significant effect on the underwater configuration curve and tension distribution; the normal drag coefficient mainly affects the underwater configuration curve, and the tangential drag coefficient mainly affects the tension distribution; different types of ocean currents have different effects and the tension of the linear current is greater than the logarithmic current.

Mobile crowd sensing (MCS), as a novel paradigm that coordinates a crowd of distributed devices to complete a whole sensing task, has attracted tremendous attention. While providing an effective and practical approach for sensing in largescale mobile scenes, the existing works on MCS suffer from a risk of privacy leakage because user data needs to be gathered in the cloud for processing and analysis. Federated Learning (FL) is a promising alternative as it can leverage mobile devices to accomplish a large learning task without centrally collecting the user data. However, incorporating FL into MCS is a non-trivial task due to the following reasons: 1) the data quality of mobile devices is often unreliable, especially in the context of crowd sensing; 2) the existing incentive mechanism in MCS may not work due to the lack of access to the user data. To address the problem, we propose a privacy-preserving mobile crowd sensing system based on Federated Learning with unreliable user data (called F-Sense). We analyze the key issues of sensing tasks, and further design an incentive mechanism to reward and motivate participants. Moreover, we explore to construct a federated quality model of user data in order to improve the data quality and obtain better training results for sensing tasks. Extensive simulation results show that F-Sense achieves privacy-preserving crowd sensing and the developed incentive mechanism can improve the task efficiency by encouraging local training at mobile devices.

Positron Emission Tomography (PET) imaging with 11C-Acetate (ACE) is regularly used in cardiovascular and in cancer imaging. In the earlier stages of ACE developments, it has been mainly used for hepatocellular carcinoma, prostate cancer, and myocardial oxygen consumption. The previous studies compared the advantage of ACE with 18F-Fluorodeoxyglucose (18F-FDG) imaging using Standard Uptake Value (SUV) and the tissue-to-blood ratio (TBR) method. The current study proposes the application of dynamic ACE PET imaging in monitoring the early response to cancer treatment. We conducted two dynamic ACE PET scans on two patients suffering from Head and Neck Cancer (HNC) (Squamous Cell Carcinoma) in the base of the tongue. Pre-treatment dynamic ACE and static 18F-FDG PET were conducted before initiation of the treatment, and the second ACE dynamic scan was performed after four weeks of radiotherapy (after 35 Gy). We applied the two-tissue compartment model to represent the kinetics of ACE in HNC. The results showed a reduction in tumor volume by more than 50% compared to the initial volume in patient-1. Besides, patient-2 has displayed a more reduced tumor volume after 4 weeks of treatment. Compartmental modeling parameter k2 increased after radiotherapy dose in both patients. This increase of k2 could reflect the reoxygenation process inside the tumor, and it can reflect the early treatment response. In conclusion, ACE could predict the early changes in the tumor perfusion and the oxidative metabolism to optimally adjust the treatment.

Affine registration can fit the non-rigid deformation of slices effectively, and it is widely used in volume reconstruction of biological tissue. But most of the existing affine registration methods are registered in a given sequence, which results in the accumulation of errors. In this paper, a global optimized affine registration method is proposed, which can be used in volume reconstruction. To eliminate the cumulative error, the affine transformation of all images is estimated simultaneously based on an energy function. A soft penalty on affine transformation is added to restrict the shearing of images. Experiments show that our method provides a more reliable registration result compared with sequential affine registration. It can solve the problems caused by the accumulation of errors. The registration result fits the deformation of slices well and preserves the rigidity of images.

The last decade has witnessed the rapid development of Internet of Things (IoT). The IoT applications are becoming more and more computation-intensive and latency-sensitive, which pose severe challenges for the resource-constrained IoT devices. To empower the computational ability of the IoT systems, edge computing emerges as a promising approach which allows the resource-constrained devices to offload their tasks to edge servers. A major challenge, which has been overlooked by most existing works on task offloading, is the dependencies among tasks and subtasks, which can have a significant impact on the offloading decisions. Besides, the existing works often consider offloading tasks to specific edge servers, which may underutilize the edge resources in the ultra-dense edge networks. In this paper, we investigate the problem of dependency-aware task offloading in ultra-dense edge networks. Specifically, we explicitly analyze the task dependency as directed acyclic graphs (DAGs) and establish full parallelism between edge servers and IoT devices. We further formulate task offloading as a joint optimization problem for minimizing both task latency and energy consumption. We prove the problem is NP-hard and propose a heuristic algorithm, which guarantees the dependency among subtasks and improves the task efficiency. Simulation experiments demonstrate that the proposed work can effectively reduce the task latency in ultra-dense edge networks.

Network Function Virtualization (NFV) transforms traditional network functions (NFs) into software through the virtualization technology, enabling easy deployment and management of NFs and improving the network flexibility. The flexibility, however, brings new challenges to network operators because of the massive deployable positions, complex NF relationships and varying network traffic. In this paper, we analyze the essential causes of the traffic dynamics and their impacts on computing and bandwidth resources regarding different NFs. We formulate the Service Function Chaining (SFC) embedding problem (which is NP-hard) and propose an intuitive solution for service function chaining for several typical cases. Motivated by the case studies, we further propose a general heuristic algorithm to solve the SFC embedding problem. Compared to the existing works, the proposed solution considers both computing and bandwidth resources and thus can achieve more efficient SFC orchestration. The extensive simulation evaluation results show that the proposed solution significantly improves the resource utilization compared to the existing works.

The practical approach to recognizing a description logic predicate in a natural language sentence is presented. The approach is based on using the open source tool – Link Grammar Parser in combination with Protege-OWL API. Machine learning method of recognition based on naive Bayes allows distinguishing kind of previously learned binary predicate and roles of its attributes. The predicate recognition system had been included into ontology learning module of the CROCUS application.

With the various emerging innovative applications, the Internet-of-Things (IoT) systems are expected to fulfill more computation-intensive and latency-sensitive sensing and computational tasks, which pose huge challenges for the IoT devices with limited computational ability and battery capacity. To address this problem, edge computing is a promising architecture where the IoT devices can offload their tasks to the edge servers. Current works on task offloading often overlook the unique task topologies and schedules from the IoT devices, leading to degraded performance and underutilization of the edge resources. In this paper, we investigate the problem of fine-grained task offloading in edge computing for low-power IoT systems. By explicitly considering 1) the topology/schedules of the IoT tasks, 2) the heterogeneous resources on edge servers and 3) the wireless interference in the multi-access edge networks, we propose a lightweight yet efficient offloading scheme for multi-user Edge systems, which offloads the most appropriate IoT tasks/subtasks to edge servers such that the expected execution time is minimized. Both centralized and distributed algorithms are devised in both sparse and dense network scenarios. We conduct extensive simulation experiments and the results show that the proposed offloading algorithms can effectively reduce the end-to-end task execution time and improve the resource utilization of the edge servers.

Pilot workload monitoring plays an important role in aviation safety. Heart rate, heart rate variability, and respiration rate have been shown to correlate with pilot workload and can be measured using electrocardiograms (ECG). Traditional machine learning approaches rely on manually extracting features from ECG, which is a difficult and time-consuming process. Recent years witnessed the success of deep neural networks, especially deep convolutional neural networks (CNNs), in computer vision and related domains; however, the application of deep CNNs onto the ECG data faces challenges on both data insufficiency and lack of tailored CNN architectures. With a small training set, this work proposes the use of transfer learning with pre-trained deep CNNs for the prediction of pilot workload. Two ECG-derived visual representations, spectrograms and scalograms, are compared for their performance on the prediction. Experimental results indicate that the scalograms perform better (at 51.35%) than spectrograms (at 45.85%) in predicting three levels of pilot workload. With the scalograms, using the pre-trained deep CNNs as "off-the-shelf" feature extractors yields better performance than fine-tuning the deep CNNs (at 42.44%) for the ECG data. The deep features are visualized using dimension reduction with t-sne., Medical Measurements and Applications (MeMeA), June 26-28, 2019, Istanbul, Turkey, Series: IEEE International Symposium on Medical Measurements and Applications (MeMeA)

In data science applications, multi-modal and multidimensional nature of data leads to various types of manual features and tailored classifiers to specific problems. Traditional feature engineering demands expertise and the performance from integrated classifiers has reached a plateau after years of development. Emerging technologies such as deep learning make a great performance leap after transition from feature engineering to architecture engineering. However, putting this technology to practical use faces numerous challenges. First, it is known that the success of deep learning is built upon large amount of training data, but most data science problems do not have sufficient training data. In addition, designing the architecture of deep models needs substantial expertise. Moreover, existing deep neural networks have limitations on the type of inputs they can work on, and their output is limited in the type of problems they can solve. To address these challenges, we propose a unified deep learning framework that suits data in multiple modalities and dimensions, and prove that using the proposed framework yields better performance than traditional machine learning approaches. Besides the examples used for validation, the proposed framework demonstrates potential in becoming a foundation, upon which more wide-spread data science problems can benefit from using deep learning., 2019 IEEE International Symposium on Medical Measurements and Applications (MeMeA), June 26-28, 2019, Istanbul, Turkey

Under the new media era, micro video is an innovative force for campus culture propaganda. Micro-video can be effectively integrated into the campus culture propaganda and construction, and subtly play the role of moral education. However, the application of micro video requires effective guidance to play a positive function, otherwise it is counterproductive. This article will focus on how to make micro-videos an effective way to strengthen campus culture propaganda based on its artistic form and characteristics, from its application development, public opinion guidance and platform mining.

Human identification is of great importance for personalized and ubiquitous computing services. As the rapid and ubiquitous deployment of wireless access, the data of channel state information (CSI) gathered from wireless networks has become a useful way for recognizing human activities. In this paper, we investigate the problem of human identification based on biometrics using the wireless big data of CSI. The rationale is that human's behavioral movements cause unique impacts on wireless CSI, which can be used to recognize the corresponding biometrics and further identify different persons in a nonintrusive manner. We first propose BioID, a general framework for biometric identification with CSI. Based on the framework, we further devise a novel identification scheme based on CSI data of lip motions. By incorporating the CSI characteristics of lip motions into noise removal and feature extraction, the proposed scheme is able to identify different users. The experimental results show that the proposed scheme can achieve accurate human identification (accuracy >90%). Finally, we discuss several open issues and challenges for future direction.

To support real-time edge computing in urban area, a number of edge servers are deployed for end devices to provide computing and storage services. Considering the massive and diverse computing requests from end devices, Network Function Virtualization (NFV) is employed for flexible service management at the edge servers. In this paper, we investigate an essential problem in NFV for large-scale edge computing – Virtual Network Function Embedding (VNFE) problem, which is NP-hard. While backup is an important means to improve the availability, we focus on the VNFE problem with backup and jointly consider the impacts of network functions/facilities. The problem is formalized as a non-linear 0-1 programming problem. We then propose a heuristic embedding algorithm based on the impacts estimation for both network functions and facilities, in which the embedding of critical Virtual Network Functions (VNFs) and servers are prioritized. We implement the algorithm and conduct extensive simulation experiments. The results show that the proposed work can greatly improve availability and resource utilization for large-scale edge computing.

The design and implementation of a highly integrated and reliable single board 8-channel transceiver at 3.5GHz is developed for massive MIMO system in 5G wireless communications. Based on this board, large scale 2D arrays, such as ${8\text{x}8}$ array, can be easily formed. On each board, 8 transmitting channels and 8 receiving channels in TDD mode are integrated and each Tx/Rx channel shares one antenna element. This transceiver can provide a verification platform for digital beamforming algorithm.

This paper presents the design and implementation of a multi-channel frequency synthesizer for 5G massive MIMO systems of 3.4–3.6 GHz. This frequency synthesizer is based on a microwave wideband synthesizer chip with integrated VCO. Several considerations for balancing the performance of phase noise, spurs, frequency resolution and locking time are discussed. The proper phase detector frequency, loop filter, RF and power filtering circuit are designed. To apply this frequency synthesizer to 64-channel massive MIMO systems with local frequency of 3.875 GHz, an 8-way Wilkinson power splitter with amplifier modules is also designed. According to this structure, the frequency synthesizer with excellent performance is developed. The experimental results indicate that the output of frequency synthesizer is with low phase noise (

The goal of connectomics research is to manifest the mechanisms and functions of neural system by using electron microscopy (EM). One of the biggest challenges in connectomic reconstruction is developing reliable neuronal membranes segmentation method to reduce the burden on manual neurite labeling and validation. In this paper, we put forward an effective deep learning approach to realize neuronal membranes segmentation in EM image stacks, which utilizes spatially efficient residual network and multilevel representations of contextual cues to achieve accurate segmentation performance. Furthermore, multicut is used as post-processing to optimize the outputs of network. Experimental results on the public dataset of ISBI 2012 EM Segmentation Challenge demonstrate the effectiveness of our approach in neuronal membranes segmentation. Our method now ranks top 3 among 88 teams and yields 0.98356 Rand Score as well as 0.99063 Information Score, which outperforms most of state-of-the-art methods.

In this paper, we propose a novel non-iterative algorithm to simultaneously estimate optimal rigid transformation for serial section images, which is a key component in volume reconstruction of serial sections of biological tissue. In order to avoid error accumulation and propagation caused by current algorithms, we add extra condition that the position of the first and the last section images should remain unchanged. This constrained simultaneous registration problem has not been solved before. Our algorithm method is non-iterative, it can simultaneously compute rigid transformation for a large number of serial section images in a short time. We prove that our algorithm gets optimal solution under ideal condition. And we test our algorithm with synthetic data and real data to verify our algorithm's effectiveness., Comment: appears in IEEE International Symposium on Biomedical Imaging 2018 (ISBI 2018)