Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-10-11 Cooperative journals: 《计算机应用研究》
Abstract: The network traffic has sudden changes and dependencies in real-time, and the traditional network traffic prediction model has the disadvantages of weak generalization ability and low prediction accuracy. In order to solve this problem, this paper developed燼 network traffic prediction model based on the Long Short-Term Memory (LSTM) recurrent neural network. Firstly, improved resampling process of Particle Filter (PF) with distance comparison and optimized combination strategy. Then, built a PF-LSTM network model to accurately predict the network traffic. Model training with improved PF algorithm to improved its training rate and overcame the shortcomings of convergence in the traditional LSTM network. Finally, predicted network traffic through proposed models. The experimental results show that the model has better prediction accuracy and convergence efficiency compared with the traditional LSTM model, which can describe the trend of changes better in network traffic.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract: Object detection is the basis of intelligent analysis, however, in the scene of illumination, shadow and clutter background, the problems of object misjudgment and unreasonable clustering is often appeared. Aiming at the above problems, propose a HVS-based object detection algorithm, which can optimize the error judgment and segmentation, and then accorded to the tracking characteristic of HVS and the continuity of object movement, combine the detection results of adjacent frames to achieve completely and accurately extraction object area. Finally, the simulation experiment based on the actual acquisition videos show that the proposed algorithm is more accurate and have good effect and in complex background.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract: Universal filtered multi-carrier (UFMC) technology is a candidate for 5G , without cyclic redundancy (CP) in transmission and inter-symbol interference (ISI) and inter-carrier interference (ICI) will occur under multi-path fading channels. To solve this problem, an equalization algorithm based on parallel interfere was proposed. Firstly, according to the analysis, the mathematical expression of UFMC system interference under multi-path channel was obtained. Secondly , after using zero-forcing equalization, the data outside the reliable interval was used to approximate the reconstruction of adjacent carriers and inter-symbol interference according to the residual interference expression. Finally, appling iterative interference cancellation to each carrier in parallel. Simulation experiments show that under multipath channel, the algorithm can reduce the bit error rate to a certain extent and improve the UFMC system performance.