Editorial:

Following the recent advances of machine learning such as deep learning algorithms and neural networks, new generation of applied artificial intelligence (AI) technologies have been used to improve intelligent communications network and its applications. New AI algorithms have been applied in various applications of information networks, such as multiple-input multiple-output (MIMO) communication networks, intelligent sensor networks, integrated satellite-terrestrial networks, Unmanned Aerial Vehicle (UAV) communication networks and so on. These new techniques have brought great changes in the field of intelligent communications network and broaden new possibilities of AI applications.

This special issue features six selected papers with high quality. The first article, “Research on anti-jamming algorithm of massive MIMO communication system based on multi-user game theory”, authored by Mingxiang Guan, Zhou Wu, Yang WeiGuo, Guo bin, Xuemei Cao and Hanying Chen, proposed anti-jamming algorithm for massive MIMO communication system. Combined with the game theory of artificial intelligence algorithm, the equipment in the communication network is modeled as the players participating in the competition in the game theory, and the array factor of the base station transmitting antenna array is adjusted according to the channel conditions and service requirements of different equipment. In order to make the Nash equilibrium point converge, an array factor update algorithm is designed.

The second article titled “3D Convolutional Neural Network for Human Behavior Analysis in intelligent sensor network” proposed a 3D max residual feature map convolution network (3D-MRCNN). The model can learn the spatial and temporal features of video data at the same time to solve the problems of network degradation and gradient disappearance. The output of the model is combined with support vector machine classifier (SVM) to improve the recognition accuracy of human behavior in video. In this method, the frame images of the input video is preprocessed by two-dimensional convolution. A learning network including 3D-max feature map (3D-MFM) and residual structure is established after the convolution splitting is completed. Finally, the output vectors corresponding to two different inputs are connected and fused into SVM to complete the human behavior classification and prediction.

In the next article with the title “Cascade Forward Artificial Neural Network based Behavioral Predicting Approach for the Integrated Satellite-terrestrial Networks”, the authors proposed a multi-step prediction approach based on a cascaded forward artificial neural network to predict user behavior in the cognitive satellite wireless network. The prediction result can help the base station in the cognitive network to schedule the dynamic access process of the cognitive users, and reduce the interference caused by the cognitive user to the authorized users.

Aiming at the poor QoS of the cell edge users in the single cell scenario, the fourth article titled “Multi-UAV Collaborative Wireless Communication Networks for Single Cell Edge Users” designs a multi-UAV system in order to provide wireless communication services for cell edge users. By considering both the resource allocation and the UAV trajectory optimization, a three-step iterative algorithm is proposed to deal with the mixed integer non-convex problem in a simplified way to obtain an approximately optimal solution. The scheme proposed in this article can significantly improve the QoS metrics for cell edge users.

Motivated by improvement of convergence rate and throughput performance, the fifth article, “High-Speed VLSI Implementation of an Improved Parallel Delayed LMS Algorithm” develops a systematic high-speed VLSI implementation of the adaptive filter based on the improved 2-parallel delayed LMS (PDLMS) algorithm. The proposed design uses a novel hardware-efficient architecture for weight updating based on parallel adaptive 2-by-2 algorithm. Compared with the conventional filter structure, the parallel filter has higher throughput rate and lower power dissipation. To improve the convergent characteristic of the adaptive digital filter, we have selected one branch from two weight update branches which has better system performance. The fine-grained arithmetic operation unit and the retiming technology are employed to reduce the delay of critical path effectively. From the ASIC synthesis results we find that the proposed architecture of an 8-tap filter has nearly 24% less power and nearly 18% less area-delay-product (ADP) than the best existing structure. Thus it can be seen that the proposed design has the important practice instruction significance.

The last article titled “Noise Robust Automatic Scoring Based on Deep Neural Network Acoustic Models with Lattice-Free MMI and Factorized Adaptation” novel noise robust automatic scoring methods for L2 speaking tests based on Deep Neural Network (DNN) models with lattice-free Maximum Mutual Information (MMI) and factorized adaptation were proposed. Noise robust Goodness of Pronunciation (GOP) algorithms using lattice free MMI were implemented to improve the reliability of automatic scoring for L2 speaking tests through better utilizing sequential training power of lattice free MMI models. Factorized adaptation for DNN acoustic models was introduced to further improve performances of the proposed GOP scores in real speaking test environments by categorizing factors that cause mismatches between acoustic models and test data. Experimental results show that the proposed methods are noise robust and outperform conventional methods in assessment for speaking tests in real classroom environments.