Search

The advent of depth sensors opens up new opportunities for human action recognition by providing depth information. The main purpose of this paper is to present an effective method for human action recognition from depth images. A multilevel frame select sampling (MFSS) method are proposed to generate three levels of temporal samples from the input depth sequences first. Then, the proposed motion and static mapping (MSM) method is used to obtain the representation of MFSS sequences. After that, this paper exploits the block-based LBP feature extraction approach to extract features information from the MSM. Finally, the fisher kernel representation is applied to aggregate the block features, which is then combined with the kernel-based extreme learning machine classifier. The developed framework is evaluated on three public datasets captured by depth cameras. The experimental results demonstrate the great performance compared with the existing approaches.

A software-defined network (SDN) is a new paradigm that separates control and forwarding. With the growth of SDN deployment scale, the centralized control plane becomes a bottleneck restricting the expansion of SDN networks. To avoid the problem of limited processing power in the wide area network (WAN), we focus on the multi-controller deployment in hierarchical architecture for software-defined WAN (SDWAN). The hierarchical architecture divides the control plane into multiple levels. The root controller has a global view. The domain controller is only responsible for the control in local networks. In this paper, we use the improved Louvain algorithm to discover the hierarchical community structure. Considering the load balancing and request delay in each partition and reliability of the control plane, we formulate the modeling and propose a hierarchical multi-controller deployment algorithm for the scalability of the SDWAN. The simulation results show that compared with the traditional algorithms, the proposed algorithm not only reduces the average request delay to a certain extent and effectively improves the performance of load balancing in each partition, but also efficiently enhances the reliability of the control plane in the SDWAN and achieves the good effect on the execution efficiency.

In actual line-commutated converter based high- voltage direct-current (LCC-HVDC) projects, multiple HVDC transmission lines are usually installed on the same tower to save transmission corridors and reduce construction costs. However, electromagnetic coupling effect between adjacent lines can cause commutation failure (CF) in the non-fault poles when a line or pole fails. Hence, this paper deeply investigates the mechanism and the prevention strategy of CFs caused by line coupling effect in single- and double-circuit HVDC systems. The influencing factors of line coupling effect and the coupling characteristics under fault conditions are analyzed first based on the modeling of parallel transmission lines and the traveling wave theory. Then the mechanism of CF caused by line coupling effect is investigated with the consideration of different fault conditions, different line structures, and different system operation modes. Based on the above analyses, prevention strategies for CFs of non-fault poles in single- and double-circuit systems are proposed. Finally, the simulation results verify the correctness of the theoretical analyses and the effectiveness of the proposed control strategies. [ABSTRACT FROM AUTHOR]

With the emergence of the hybrid multi-infeed high voltage direct current (HMIDC) system, there is complex dynamic interaction between the AC/DC system and converter stations, while the interaction mechanism of the HMIDC system is unclear. For this purpose, this paper establishes the HMIDC linearized model based on the motion equation concept. Then, after decoupling and reconstructing the interaction path of the HMIDC system, the damping coefficients are derived. The AC system strength and voltage source converter (VSC) terminal voltage controller parameters are selected to analyze the impacts on damping coefficients, and the interaction mechanism of the HMIDC system is analyzed. In order to solve the potential instability issue, an additional damping controller in the VSC terminal voltage loop is proposed to enhance the voltage stability of HMIDC system, which is realized by adjusting the reactive power/amplitude coupling damping coefficient. Finally, the RT-LAB-based hardware-in-the-loop experiments verify the accuracy of the interaction mechanism analysis of the HMIDC system and the effectiveness of the proposed controller. [ABSTRACT FROM AUTHOR]

Deep learning (DL) has exhibited huge potentials for hyperspectral image (HSI) classification due to its powerful nonlinear modeling and end-to-end optimization characteristics. Although the superior performance of DL-based methods has been witnessed, some limitations can still be found. On the one hand, existing DL frameworks usually resorted to first-order statistical features, whereas they rarely considered second-order or higher order statistical features. On the other hand, the optimization of complex hyperparameters (e.g., the layer number and convolutional kernel size) is time-consuming and a very tough task, making the designed DL framework unexplainable. To overcome these challenges, we propose a novel attention-based second-order pooling network (A-SPN). First, a first-order feature operator is designed to model the spectral–spatial information of HSI. Second, an attention-based second-order pooling (A-SOP) operator is designed to model discriminative and representative features. Finally, a fully connected layer with softmax loss is used for classification. The proposed framework can obtain second-order statistical features in an end-to-end manner. In addition, A-SPN is free of complex hyperparameters tuning, making it more explainable and easily equipped for classification tasks. Experimental results based on three common hyperspectral data sets demonstrate that A-SPN outperforms other traditional and state-of-the-art DL-based HSI classification methods in terms of generalization performance with limited training samples, classification accuracy, convergence rate, and computational complexity. [ABSTRACT FROM AUTHOR]

In a line-commutated converter based high-voltage direct-current (LCC-HVDC) system, the inverter control system usually includes constant extinction angle (CEA) controller, constant current (CC) controller, current error controller (CEC), and voltage-dependent current order limiter (VDCOL). However, the interaction influence of multiple controllers may cause repetitive commutation failure (CF). For this purpose, this paper deeply investigates the mechanism and the influencing factors of repetitive CF caused by controller interaction. It is found that the lower limit value of CEA controller can affect the control switching process and thus the probability of repetitive CF. Besides, the minimum gamma measurement (MGM) unit prevents the system from acquiring the real extinction angle in time, which results in the DC current fluctuation during the recovery process. Moreover, the probability of control-induced repetitive CF is much higher when the HVDC inverter is connected to a weak AC system or the AC fault is not serious but not cleared in time. Based on the above analyses, an improved control method is proposed for recovery performance improvement, which is mainly realized by modifying the lower and upper limits of the CEA controller during recovery. Finally, the simulation and experimental results verify the theoretical analyses and the proposed control method. [ABSTRACT FROM AUTHOR]

Load converter operating in the weak grid potentially suffers harmonic oscillation problems due to the impedance interaction. In this article, considering dc-link voltage dynamics and frequency-coupling effects, the precise sequence impedance models are derived for the traditional load converter (TLC) and the load virtual synchronous machine (LVSM). Based on the derived models, their sequence impedance characteristics are analyzed and compared. The analysis shows that the positive-sequence impedance of the LVSM is generally inductive in the middle- and low-frequency bands, which is basically consistent with the grid impedance characteristics. In contrast, the positive-sequence impedance of the TLC is negative-resistive-capacitive (i.e., phase angle is between -180° and -90°) in the middle- and low-frequency bands and its amplitude-frequency curve is easy to intersect with the amplitude-frequency curve of the grid impedance, which may lead to harmonic oscillation in the weak grid. Furthermore, based on the derived model and Nyquist stability criterion, the effects of the grid impedance and load power on the stability are analyzed for TLC and LVSM, which indicates that the LVSM has better stability in the weak grid than the TLC. Finally, the experimental results validate the correctness of the theoretical analyses. [ABSTRACT FROM AUTHOR]

This letter analyzes the mechanism of commutation failures (CFs) caused by sending end AC faults. It is founded that the dc impulse current after fault clearance is the direct reason of CF occurring under nonserious sending end AC faults, and the controller interaction is the main reason of CF occurring under severe sending end AC faults. Moreover, a prevention strategy is proposed based on the revealed mechanism, and the validity is verified by the simulation results. [ABSTRACT FROM AUTHOR]

Load converters with inertial support are used to enhance the system inertia. However, these converters may fail to operate stably in the weak grid. In this article, sequence impedance modeling and stability comparison analysis are proposed for the load virtual synchronous machine (LVSM) and the load converter with virtual inertia control (LCVIC). First, the universal sequence impedance formulas are derived by considering dc-link voltage dynamics and frequency-coupling effects. Then, the precise sequence impedance models are built for impedance characteristics analysis of LVSM and LCVIC. The analysis shows that the positive-sequence impedance of LVSM is generally inductive in the middle-frequency band, which is basically consistent with the grid impedance. In contrast, the positive-sequence impedance of LCVIC is negative-resistive-capacitive (i.e., phase angle is between −180° and −90°) in the middle-frequency band, which may lead to harmonic oscillation in the weak grid. Furthermore, the effects of grid impedance, load power, and other control parameters on the stability are analyzed for the two load converters based on the derived model and Nyquist stability criterion. The analysis results reveal that LVSM has better stability than LCVIC in the weak grid. Finally, the experimental results validate the correctness of the theoretical analyses. [ABSTRACT FROM AUTHOR]

Multimedia services are predominant in current wireless networks and are becoming ubiquitous in the upcoming 5G era in which the video quality of experience (QoE) is a fundamental metric. However, no widely accepted QoE model exists due to its subjective nature. This paper proposes a framework for quantifying the QoE of multimedia content based on the facial expression approach, which can directly reflect the end users' intrinsic attitudes toward the services. To achieve this objective, a face database is established containing over one thousand videos and serves as a dataset for the subsequent experience mining. The semi-supervised clustering method proposed in this paper is applied to calculate the video experience scores and achieves 8% higher average test accuracy than other prevailing methods. Extensive experimental results show that our approach can accurately reveal the user's experience toward video contents and is expected to become a valid and useful QoE model. [ABSTRACT FROM AUTHOR]