In order to improve the accuracy of video anomaly behavior detection, a new detection method based on generative adversarial networks for video foreground areas is proposed. First, the foreground and background masks of the ground truth video frame are extracted, to determine the foreground areas of the output video frames from generative adversarial networks. For the foreground areas under consideration, the foreground area peak signal-to-noise ratio (F-PSNR) is used to calculate the detection score of anomaly behaviors. The experimental results show that the proposed method can effectively improve the detection accuracy of video anomaly behaviors with a reduced detection time for the Avenue dataset, UCSD-Ped1 dataset and UCSD-Ped2 dataset.

The energy-efficient caching strategy and computation offloading design of mobile edge computing (MEC) faces the double randomness challenges, which require to adapt to the time-varying wireless channel state and the dynamic task arrivals. This paper investigates a cache-enabled mobile edge computing system with dynamic tasks arriving at multiple wireless devices. By minimizing the system weighted sum energy over multiple time slots, we optimize the AP's task caching decision and MEC execution, the wireless devices’ local computing and tasks offloading under the caching capacity and computation causality, and the computation deadline constraints. The branch-and-bound (BnB) method is first presented to obtain the globally optimal solution to define the lower bound for practical schemes. Then, a relaxation-based scheme is proposed to efficiently achieve a near-optimal solution. Numerical results show that the proposed relaxation-based scheme achieves a closer performance to the optimal BnB scheme when compared to the benchmark schemes.

A wireless powered over-the-air computation (AirComp) system is studied, where one separately-located energy transmitter (ET) is deployed to charge multiple low-power sensors simultaneously via energy beamforming, and these sensors rely on the harvested energy for sequential data sensing and functional computation along with the access point (AP) . A harvest-then-sense-and-transmit protocol is considered. Under this system setup, an energy-efficient AirComp design is pursued to minimize the transmit energy of the ET, subject to the sensor energy harvesting constraints and the computational mean squared error (MSE) constraints. The energy beamforming vectors of the ET, the receive beamforming vectors of the AP, and the transmit coefficients at the sensors are jointly optimized. Due to the complicated variable coupling, the resultant energy minimization problem is non-convex. As such, an alternating optimization method is presented to obtain a near-optimal design solution in an iteration manner. Numerical results are provided to show the fast convergence performance and the merit of the proposed design solution.

In order to make full use of the computing resources of Mobile Edge Computing (MEC) system, this paper designs an optimization scheme of collaboration between two MEC servers and joint computing and communication resources. The scheme proposes the optimization problem of dual-server collaborative multi-user task calculation, where the weighted sum between system computing delay and user energy consumption is minized. The multi-user computing unloaded transmitting power and task segmentation are optimized in the proposed scheme. A joint design scheme with low computational complexity is proposed. The original problem is decoupled into two sub-problems of computational offload optimization and computational task segmentation design, both of which can be solved by interior point method and simplex method respectively. The simulation results show that the system performance of the proposed scheme is better than the existing heuristic benchmark algorithm scheme. And the joint optimization algorithm scheme can get the similar system performance as compared with the basic scheme of the optimal Lagrange multiplier method with less computation time.

In order to solve the micro battery energy problem of terminal Internet of Things user devices in computation-intensive applications, green renewable energy harvesting technology and double reconfigurable intelligent surfaces(RIS) technology enabling edge computing were studied and a green Internet of Things edge computing system assisted by double-RIS was constructed, effectively extending the computing life of terminal Internet of Things user devices and improving system computation capacity. Firstly, a multi-user cascade fading channel model assisted by double-RIS was established, and a multi-slot random arrival model of green renewable energy harvesting was established to model the causal constraints of energy supply and demand of Internet of Things terminal devices. Secondly, based on the maximization of system computation capacity, the joint optimization design problem of terminal local computing rate, edge computing offloading power and phase shift of RISs was modeled. This design problem belongs to a class of complex non-convex optimization problem. To this end, lightweight multi-stage optimization technology was adopted to rapidly and iteratively design variables such as local computation, computation offloading and phase shift of RISs, etc, to complete the design of green Internet of Things edge computing system. The experimental results show that the performance gains of the proposed scheme are better than the existing benchmark schemes, and the proposed scheme is equivalent to the scheme based on semidefinite relaxation algorithm under less system computing time.

The reconfigurable intelligent surface (RIS) -assisted integrated sensing and communication (ISAC) system is a key technology for future 6G networks. A cooperative dual-active RIS-assisted ISAC system is studied, utilizing dual-reflection links to achieve both communication between a base station and multiple users in non-line-of-sight areas and multi-target sensing tasks. To maximize the minimum sensing beampattern gain, the communication quality of service, sensing mean square cross-correlation, and power constraints for both the base station and active RIS were modeled, forming a joint optimization problem for key parameters including the communication beamforming and sensing covariance matrix at the base station, and the reflection coefficient matrix and amplification power allocation of the active RIS. Due to the strong coupling between the beamforming vectors, sensing covariance matrix at the base station, and the reflection coefficient matrices of the two active RISs, this joint optimization problem is non-convex and highly complex to solve. Therefore, a distributed iterative suboptimal algorithm was designed to decouple the joint optimization into three subproblems: optimizing the communication beamforming, sensing covariance matrix, and active RIS amplification power allocation at the base station, and optimizing the reflection coefficient matrices at both active RISs. The simulation results demonstrate that the proposed distributed optimization algorithm exhibits good convergence performance. Compared with other schemes, optimizing the amplification power allocation between the two active RIS units can improve the beampattern gain by approximately 2 dB under the same system power budget and number of RIS elements. This leads to higher power utilization efficiency.

In order to solve the challenges of video anomaly behavior detection, a new detection algorithm model is proposed based on masked convolution and external attention mechanism convolutional neural network. On the one hand, by masked convolution, it can restrict the effective regions of convolution to make the neural network efficiently learn, so as to model the positive moderate that is effective in the feathers of normal behavior. On the other hand, by combining with lightweight external attention mechanisms, the modeling quality of interested regions is improved. Convolutional neural network in reconstruction or predictive architectures is added in feature reconstruction loss to improve the detection accuracy of video abnormal behavior detection. The experimental results show that the proposed method can effectively improve the detection performance of video anomaly behaviors by 2.86%, 2.54% and 0.56% on the Avenue dataset, UCSD-Ped1 dataset and UCSD-Ped2 dataset, respectively.