As the key technology of fifth generation (5G), 3-dimensional (3D) massive multi-input and multi-output (MIMO) is expected to be widely used in small cell network (SCN). In this paper, in order to investigated the tradeoff between limited size in SCN and the capacity gain from increasing antenna elements，the spatial performances of 3D massive MIMO based on a MIMO channel measurements at 6 GHz in urban microcell (UMi) scenario are studied. Enormous channel impulse responses (CIR) are collected and reconstructed, which enables us to present comparative results of the capacity and the eigenvalue spread (ES). Furthermore, the impacts of antenna element number and spacing on system performance are investigated, i.e., 32, 64, 128 elements are selected from the 512 transmitter (Tx) array with elevation interval spacing being 0.5, 1 and 2 wavelengths for each. Interestingly, the capacity gap can be obviously observed on the comparison between the 1 and 2 wavelength antenna spacing cases, which implies that correlation cannot be ignored when the antenna spacing is larger than 1 wavelength when massive antennas are equipped. The contrast results show that the capacities are enlarged with the increasing of antenna elements number, and larger antenna spacing will lead to higher channel capacity as expected. However, the capacity gains brought by the increasing of antenna spacing will descend to certain degrees as the antenna number increases. Collectively, these results will provide further insights into 3D massive MIMO utilization.

With the intensive deployment of users and the drastic increase of traffic load, a millimeter wave (mmWave) back-haul network was widely investigated. A typical mmWave back-haul network consists of the macro base station (MBS) and the small base stations (SBSs). How to efficiently associate users with the MBS and the SBSs for load balancing is a key issue in the network. By adding a virtual power bias to the SBSs, more users can access to the SBSs to share the load of the MBS. The bias values shall be set reasonably to guarantee the back-haul efficiency and the quality of service (QoS). An improved Q-learning algorithm is proposed to effectively adjust the bias value for each SBS. In the proposed algorithm, each SBS becomes an agent with independent learning and can achieve the best behavior, namely the optimal bias value through a series of training. Besides, an improved behavior selection mechanism is adopted to improve the learning efficiency and accelerate the convergence of the algorithm. Finally, simulations conducted in the 60 GHz band demonstrate the superior performance of the proposed algorithm in back-haul efficiency and user outage probability.

In orthogonal frequency division multiplexing/offset quadrature amplitude modulation (OFDM/OQAM) systems, the relationship between the input of the synthesis filter bank (SFB) and the output of the analysis filter bank (AFB) is much more complicated than OFDM due to its special prototype filter. By analyzing the trans-multiplexer response characteristics, an equivalent trans-multiplexer matrix is proposed to describe the relationship between the input and the output. With the equivalent matrix, the output can be easily computed using matrix multiplication with the input. Moreover, with the inverse of the equivalent trans-multiplexer matrix, imaginary interference can be eliminated using the precoding method. The simulation results show the correctness of the equivalent trans-multiplexer matrix.

Wireless ultra-dense network (UDN) is one of the important technologies to solve the burst of throughput demand in the forthcoming fifth generation (5G) cellular networks. Reusing spectrum resource for the backhaul of small base stations (SBSs) is a hotspot research because of lower cost and rapid implementation with macro base stations (MBSs) in recent years. In heterogeneous UDN, the problem of spectrum allocation for wireless backhaul is investigated. In particular, two different spectrum resource reusing strategies for wireless backhaul are proposed in heterogeneous UDN with the limited bandwidth condition. Using a stochastic geometry-based heterogeneous UDN model, the success probabilities that mobile users communicate with SBSs or MBSs are derived under two different spectrum resource reusing strategies. In addition, the network throughput’s analytical expressions and the optimal ratio of spectrum allocation are derived. Numeral results are provided to evaluate the performance of the proposed strategies at throughput. Thus, the effectiveness of the strategy that mobile users can only communicate with SBSs is validated.

In order to overcome the poor generalization ability and low accuracy of traditional network traffic prediction methods, a prediction method based on improved artificial bee colony (ABC) algorithm optimized error minimized extreme learning machine (EM-ELM) is proposed. EM-ELM has good generalization ability. But many useless neurons in EM-ELM have little influences on the final network output, and reduce the efficiency of the algorithm. Based on the EM-ELM, an improved ABC algorithm is introduced to optimize the parameters of the hidden layer nodes, decrease the number of useless neurons. Network complexity is reduced. The efficiency of the algorithm is improved. The stability and convergence property of the proposed prediction method are proved. The proposed prediction method is used in the prediction of network traffic. In the simulation, the actual collected network traffic is used as the research object. Compared with other prediction methods, the simulation results show that the proposed prediction method reduces the training time of the prediction model, decreases the number of hidden layer nodes. The proposed prediction method has higher prediction accuracy and reliable performance. At the same time, the performance indicators are improved.

The explosive increase of smart devices and mobile traffic results in heavy burden on backhaul and core network, intolerable network latency and degraded service to the end-users. As a complement to core network, edge network contributes to relieving network burden and improving user experience. To investigate the problem of optimizing the total consumption in an edge-core network, the system consumption minimization problem is formulated considering the energy consumption and delay. Given that the formulated problem is a mixed nonlinear integer programming (MNIP), a low-complexity workload allocation algorithm is subsequently proposed based on interior-point method. The proposed algorithm has an extremely short running time in practice. Finally, simulation results show that edge network can significantly complement core network with much reduced backhaul energy consumption and delay.

Edge is the intrinsic geometric structure of an image. Edge detection methods are the key technologies in the field of image processing. In this paper, a multi-scale image edge detection method is proposed to effectively extract image geometric features. A source image is decomposed into the high frequency directional sub-bands coefficients and the low frequency sub-bands coefficients by non-subampled contourlet transform (NSCT). The high frequency sub-bands coefficients are used to detect the abundant details of the image edges by the modulus maxima (MM) algorithm. The low frequency sub-band coefficients are used to detect the basic contour line of the image edges by the pulse coupled neural network (PCNN). The final edge detection image is reconstructed with detected edge information at different scales and different directional sub-bands in the NSCT domain. Experimental results demonstrate that the proposed method outperforms several state-of-art image edge detection methods in both visual effects and objective evaluation.

Non-binary low density parity check (NB-LDPC) codes are considered as preferred candidate in conditions where short/medium codeword length codes and better performance at low signal to noise ratios (SNR) are required. They have better burst error correcting performance, especially with high order Galois fields (GF). A shared comparator(SCOMP) architecture for elementary of check node (ECN)/elementary of variable node (EVN) to reduce decoder complexity is introduced because high complexity of check node (CN) and variable node (VN) prevent NB-LDPC decoder from widely applications. The decoder over GF(16) is based on the extended min-sum (EMS) algorithm. The decoder matrix is an irregular structure as it can provide better performance than regular ones. In order to provide higher throughput and increase the parallel processing efficiency，the clock which is 8 times of the system frequency is adopted in this paper to drive the CN/VN modules. The decoder complexity can be reduced by 28% from traditional decoder when shared comparator architecture is introduced. The result of synthesis software shows that the throughput can achieve 34 Mbit/s at 10 iterations. The proposed architecture can be conveniently extended to GF such as GF(64) or GF(256). Compared with previous works, the decoder proposed in this paper has better hardware efficiency for practical applications.

With the popularity of adaptive multi-rate wideband (AMR-WB) audio in mobile communication, many AMR-WB based techniques, such as a similar compression architecture to transmit secret information during the process of compression, were proposed to transmit covert messages. However, if a sender does not have the original WAV audio, the architecture cannot be used. In this paper, a new covert message method, which takes effect after WAV audio is compressed into AMR-WB speech, is proposed. This method takes advantage of algebraic codebook search. Aiming at improving speed and reducing search space, it does not perform algebraic codebook search using the optimal search algorithm, and it does not reach the positions of non-zero pulses via depth-first tree search that characterizes the energy of audio. According to the features of search methods and the codebook index construction, every track in each subframe is analyzed to find the proper positions for embedding secret information. Experimental results show that the proposed method has satisfactory capacity and simplicity regardless of compression process.

Hard competition learning has the feature that each point modifies only one cluster centroid that wins. Correspondingly, soft competition learning has the feature that each point modifies not only the cluster centroid that wins, but also many other cluster centroids near this point. A soft competition learning method is proposed. Centroid all rank distance(CARD), CARDx, and Centroid all rank distance batch K-means(CARDBK) are three clustering algorithms that adopt the soft competition learning method proposed by us. Among them the extent to which one point affects a cluster centroid depends on the distances from this point to the other nearer cluster centroids, rather than just the rank number of the distance from this point to this cluster centroid among the distances from this point to all cluster centroids. In addition, the validation experiments are carried out in order to compare the three soft competition learning algorithms CARD, CARDx, and CARDBK with several hard competition learning algorithms as well as neural gas(NG) algorithm on five data sets from different sources. Judging from the values of five performance indexes in the clustering results, this kind of soft competition learning method has better clustering effect and efficiency, and has linear scalability.

A novel asymmetrical Pi-shaped defected ground structure (DGS) with 3-interations Koch fractal curves is proposed to design a microstrip low-pass filter (LPF) with ultra-wide stop-band (SB). The proposed LPFs with a single resonator and two cascaded resonators are both designed, simulated, manufactured and measured. Simulation and experiment results demonstrate that the designed LPF has a very sharp transition band (TB) and an ultra-wide SB performance compared with the existed similar symmetrical and asymmetrical DGS. The proposed LPF with two cascaded resonators is with a compact size of 36.8 mm×24.0 mm, a very low insertion loss of less than 0.7 dB under 1.9 GHz, and a wide SB from 2.2 GHz to 8 GHz with rejection of larger than 30 dB.