Table of Content

20 December 2018 Volume 45 Issue 6

  

New method for missile positioning based on SAR scene matching

LIANG Yi, LI Qingqing, SUN Kun, DANG Yanfeng, DING Jinshan

Journal of Xidian University. 2018, 45(6):  1-6.  doi:10.3969/j.issn.1001-2400.2018.06.001

Abstract ( 911 )   PDF (603KB) ( 288 )   Save

References | Related Articles | Metrics

 During the final guidance stage, the position deviation of the missile flight trajectory is often corrected by scene matching technology, but traditional positioning methods based on Doppler information are sensitive to errors, which can not satisfy the requirement of accurate navigation. To solve  this problem, a new method for missile positioning is proposed. First, the location information on multi-feature points in the missile-borne SAR image after scene matching is obtained by building a positioning geometric model of the arbitrary point as the image coordinate system origin. Then the nonlinear equations are constructed by corresponding slant range information. The actual position of the missile is calculated by the Gaussian Newton-genetic hybrid algorithm, so that the trajectory deviation is corrected accurately.





Analysis of scattering characteristics of the one-thousand-wavelength object using the parallel IE-NDDM

SU Qin, GU Zongjing, LIN Zhongchao, ZHAO Xunwang, ZHANG Yu

Journal of Xidian University. 2018, 45(6):  7-13.  doi:10.3969/j.issn.1001-2400.2018.06.002

Abstract ( 398 )   PDF (1743KB) ( 121 )   Save

References | Related Articles | Metrics

In order to solve the scattering characteristics of an electrically-large object in the condition of limited resources fast and accurately, a parallel nonoverlapping non-conformal domain decomposition method based on integral equations is proposed. The parallel multilevel fast multipole algorithm is used to accelerate the calculation of the interior and the coupling of sub-domains, and the adaptive direction partitioning scheme is applied to improve the parallel efficiency according to its oct-tree structure. The coupling between different sub-domains is calculated in the manner of near field to avoid the storage of the mutual impedance and the memory requirement is further reduced. Numerical results prove that the proposed parallel domain decomposition method can efficiently solve the scattering problem of one thousand wavelengths.





Near-field beam synthesis algorithm for sparse array antennas

SONG Jiaqi, TAO Haihong

Journal of Xidian University. 2018, 45(6):  14-18.  doi:10.3969/j.issn.1001-2400.2018.06.003

Abstract ( 473 )   PDF (1286KB) ( 145 )   Save

References | Related Articles | Metrics

Due to the problem that the large aperture of sparse array antennas leads to most interested targets lying in the near-field, we propose a novel near-field beam synthesis method for sparse array antennas. By utilizing the propagation model of near-field electromagnetic waves, we analyze the original propagation delay expression. When the number of antenna elements and the interval bounds are fixed, this method achieves a lower peak side-lobe level of the linear sparse array via differential evolution, with the steps of the algorithm given. Computer simulation shows the effectiveness and feasibility of the algorithm by realizing near-field beam focusing and PSLL controlling under high sparseness. It is proved that the proposed method reduces the high peak side-lobe level brought by the large array aperture, and has better performance than traditional methods.







Method for automatic prediction of the development trend of an ophthalmic disease

JIANG Jiewei, LIU Xiyang, LIU Lin, WANG Shuai, YANG Haoqing, CUI Jiangtao

Journal of Xidian University. 2018, 45(6):  19-25.  doi:10.3969/j.issn.1001-2400.2018.06.004

Abstract ( 426 )   PDF (2309KB) ( 125 )   Save

References | Related Articles | Metrics

The current researches on the computer-aided diagnosis of an ophthalmic disease focus mainly on the automatic classification or grading based on the currently available images, the method for the prediction of an ophthalmic disease is scarce, and therefore, a cost-sensitive temporal sequence method is proposed to analyze and predict the development trend of an ophthalmic disease. First, the Canny edge detector operator and Hough transform are used to preprocess the slit-lamp image and obtain the lens area. Second, the residual convolutional neural network is employed to extract the high-level features from the lens area, which are then inputted into the long short term memory network to mine the inherent laws between the temporal sequence data. Finally, the cost-sensitive Softmax classifier is used to predict the development trend of an ophthalmic disease. Experimental results prove that this method has higher accuracy and sensitivity for prediction, and can simultaneously predict different sequence data with a length of 3～5.







Coding and decoding algorithms for systematic permutation codes at the Chebyshev distance

HAN Hui, MU Jianjun, JIAO Xiaopeng

Journal of Xidian University. 2018, 45(6):  26-30+98.  doi:10.3969/j.issn.1001-2400.2018.06.005

Abstract ( 342 )   PDF (439KB) ( 69 )   Save

References | Related Articles | Metrics

Since existing ［k+n,k,d］ systematic permutation codes which can correct limited-magnitude errors lack encoding and decoding procedures in the Chebyshev metric, an encoding algorithm for these systematic permutation codes is proposed by using the ranking and unranking mappings in the symmetric group and the interleaving technology of (n,M,d) permutation codes in the Chebyshev metric. Moreover, a decoding method of ［k+n,k,d］ systematic permutation codes is presented by using the the ranking and unranking mappings in the symmetric group and the projection technology for (n,M,d) permutation codes in the Chebyshev metric. The correctness of the proposed encoding and decoding methods for systematic permutation codes is illustrated by some calculation examples.





Robust energy efficiency optimization for secrecy MIMO systems with SWIPT

YANG Guangyu, QIU Hongbing

Journal of Xidian University. 2018, 45(6):  31-37.  doi:10.3969/j.issn.1001-2400.2018.06.006

Abstract ( 721 )   PDF (575KB) ( 82 )   Save

References | Related Articles | Metrics

This paper investigates the secrecy energy efficiency optimization for simultaneous wireless information and power transfer in the multiple-input-multiple-output wiretapping system which consists of  multiple eavesdroppers (Eves). In the presence of imperfect channel estimation ,the beamforming matrix, the artificial noise covariance matrix and the power splitting ratio are jointly designed to maximize the worst-case secrecy energy efficiency under the maximum transmit power and the minimum harvested energy constraints. Due to the non-convexity of this problem, the objective function is recast into an equivalent parametric problem by exploiting the fractional programming theory, and then the non-convex equivalent problem is approximated as a convex optimization problem by means of the S-Procedure and first order Taylor series  expansion. Furthermore, a two level optimization iterative algorithm is proposed based on the Dinkelbach method. Finally, simulation results are provided  to show the effectiveness and convergence of  the proposed algorithm.

ADP interference suppression technique for the parasitic transmission system

SUN Linhai, TIAN Bin, HUANG Junhua, ZHANG Yueting

Journal of Xidian University. 2018, 45(6):  38-43.  doi:10.3969/j.issn.1001-2400.2018.06.007

Abstract ( 356 )   PDF (628KB) ( 74 )   Save

References | Related Articles | Metrics




Considering the broadband interference caused by the broadcast signal to the parasitic signal in the Parasitic Transmission System, this paper provides an interference suppression technology based on Amplitude Domain Processing(ADP). With the simplified method of Continuous Polynomial Approximation for expressing mathematically the Probability Density Function(PDF), the nonlinear function derived from this PDF can be applied to nonlinear processing in the amplitude domain of the received signal to suppress interference. This technique reduces the error from inaccuracy of PDF estimation and does not need any training sequence. Simulation results indicate that the Parasitic Transmission System with this technique can effectively suppress broadband interference of the QPSK modulated signal, and that the bit error rate(BER) of the system is about 10^-3, even when the interference-to-signal ratio is 20dB.

State transition-based access control model in the UAV network

MA Zhuoran, MA Jianfeng, MIAO Yinbin, SUN Cong

Journal of Xidian University. 2018, 45(6):  44-50.  doi:10.3969/j.issn.1001-2400.2018.06.008

Abstract ( 336 )   PDF (1053KB) ( 78 )   Save

References | Related Articles | Metrics

Due to many types of resources and high mobility in an Unmanned Aerial Vehicle (UAV) network, the traditional access control mechanism cannot achieve the flexible permission management when the authorization changes and the task transfers among UAVs. Therefore, an access control model based on state transition is proposed. The state transition rules are determined by the access properties of object resources, thereby achieving the flexible permission management and subject priority when the authorization changes and resource permissions transfer. Moreover, the experiments in the UAV network environment demonstrate that the execution time for the proposed scheme almost linearly increases with the growing scale of subjects and objects, and that the scheme is feasible in an actual UAV network with complex access control.

Analysis of the miniaturization of the anti-jamming array antenna and its beamforming technology

LI Feng, WANG Ziyang, WANG Le, LI Rui, FENG Gang, YIN Yingzeng

Journal of Xidian University. 2018, 45(6):  51-56.  doi:10.3969/j.issn.1001-2400.2018.06.009

Abstract ( 670 )   PDF (560KB) ( 179 )   Save

References | Related Articles | Metrics

Based on the miniaturization of the anti-jamming array antenna, a new technique for beamforming is proposed. In the actual design, the array antenna is miniaturized by reducing the spacing between array antenna elements; the HFSS VBScript script is programmed to extract the active pattern and active impedance data of array elements; the optimization based on the differential evolution algorithm is written in Matlab; HFSS is called by Matlab for full-wave simulation analysis, and the optimal element weighting factors are obtained to achieve the pattern null of the anti-jamming array antenna. Research results show that this technique can form pattern null in the interference direction (30°,10°)very accurately, and that the null depth  increases by 32dB.





Design of a low-power low-temperature-drift LDO regulator in mobile chips

CHEN Diping, YING Tao, DONG Gang

Journal of Xidian University. 2018, 45(6):  57-62.  doi:10.3969/j.issn.1001-2400.2018.06.010

Abstract ( 568 )   PDF (1160KB) ( 168 )   Save

References | Related Articles | Metrics

To fulfill the need of low-power consumption on mobile devices, an output-capacitorless low dropout regulator is designed based on the GSMC 0.18μm CMOS technology. On the basis of the framework of a conventional low dropout regulator circuit, a constant-current source with temperature compensation instead of a feedback resistor is placed and used as the bias reference current for the reference voltage source and the error amplifier. Consequently, the quiescent power is reduced, and in the meantime, both the temperature compensation and adjustability of the output voltage are achieved. The result shows that the quiescent current is only 5.486μA under a voltage between 2.85V and 4.00V, that the temperature coefficient of the output voltage is 9.772×10^-6/℃ with a temperature between -40℃ and 85℃, and that the active layout area of this LDO is 0.12mm×0.09mm.




Computation offloading approach achieving optimization of both processing delay and energy efficiency for mobile terminals

ZHANG Wenzhu, CAO Beibei, KONG Weipeng

Journal of Xidian University. 2018, 45(6):  63-68+111.  doi:10.3969/j.issn.1001-2400.2018.06.011

Abstract ( 242 )   PDF (768KB) ( 52 )   Save

References | Related Articles | Metrics

Computation offloading is a technique to offload computation intensive applications executed on mobile terminals to remote servers. It is an effective way to expand the comprehensive processing capabilities of mobile terminals. However, for this promise to become a reality, it is necessary to find an approach to guide decision-making on computation offloading. Orienting to the LTE context, a computing offloading approach is presented in this paper, which can optimize both processing delay and energy efficiency. The approach first establishes the mobile terminal's computation offload model; then it constructs an offloading cost function based on analyzing the computation offloading parameters related; finally, it implements reasonable computation offloading by taking the processing delay and the energy consumption as constraints. An extensive numerical analysis reveals that the proposed approach achieves less processing delay and better energy efficiency simultaneously.

Particle swarm optimization method based on dynamic sub-swarms with entropy weight

LIU Daohua, HU Xiuyun, ZHAO Yansong, CUI Yushuang

Journal of Xidian University. 2018, 45(6):  69-74.  doi:10.3969/j.issn.1001-2400.2018.06.012

Abstract ( 411 )   PDF (928KB) ( 75 )   Save

References | Related Articles | Metrics

To improve the performance of particle swarm optimization, a particle swarm optimization method based on dynamic sub-swarms with entropy weight is proposed. The method of k-means is applied to obtain the number of subgroups, and during the course searching, to utilize the entropy information about other particles, optimal solution information from the subgroup searching process and those from other subgroups are used to form the entropy weight so as to adjust the inertia weight, and the entropy weight is formed by optimization information of m times iterations to adjust the global optimization solution of the particle swarm. During the fine searching, optimization information obtained from each particle swarm is used as the initial setting of the new swarm, and the iteration information about other particles is used to from the entropy weight to adjust the global optimization solution. Some traditional methods and the proposed method in this paper are compared with four classical test functions, and the results show that the method proposed in the paper has advantages of high precision and fewer iterations.

High transfer efficiency control method for AC/DC

LUO Xiaoqing, CHEN Yan, HU Rong, HE Shangping

Journal of Xidian University. 2018, 45(6):  75-79.  doi:10.3969/j.issn.1001-2400.2018.06.013

Abstract ( 570 )   PDF (1666KB) ( 110 )   Save

References | Related Articles | Metrics

To improve the transfer efficiency of the AC/DC converter system working in the DCM(Discontinuous Conduction Mode) mode, the power consumption mechanism of the system in different states in one switching period is studied and analyzed, on the basis of which the power stage model is constructed. By MATLAB simulation as well as analysis, the power loss curves of key factors are obtained. On the basis of the analysis, a high efficient constant voltage control method is proposed. The constant voltage  control strategy is realized by a 0.35μm/700V/BCD(Bipolar-CMOS-DMOS) process, and simulation and experiment show that the strategy has an excellent load transient response. The transfer efficiency when the power consumption is 5W can reach a level as high as 83%, which exceeds the newest DoE.6 standard, thus realizing the high transfer efficiency.

Beam precoding algorithm for antenna mutual coupling effect suppression

ZHONG Weizhi, XU Lei, ZHU Qiuming, CHEN Xiaomin, ZHOU Jianjiang

Journal of Xidian University. 2018, 45(6):  80-85.  doi:10.3969/j.issn.1001-2400.2018.06.014

Abstract ( 314 )   PDF (677KB) ( 121 )   Save

References | Related Articles | Metrics

A beam precoding method based on the improved orthogonal matching pursuit (OMP) algorithm is proposed for the millimeter-wave hybrid beamforming system, combined with the analysis of the antenna mutual coupling effect. This method first deduces the antenna mutual coupling matrix, on the basis of which the equivalent mutual coupling channel model is established. Second, with the goal of maximizing the transmission rate, the precoding vector is calculated by utilizing the improved OMP algorithm to obtain the optimal beam under the mutual coupling effect, when the equivalent mutual coupling channel model has been established. Simulation shows that the beam precoding method proposed in this paper can effectively suppress the beam distortion caused by the mutual coupling effect and guarantee the communication quality.

Fast and high-resolution VCO frequency calibration technique

CAI Qingsong, YANG Zhong, JIA Xiaoyun, QIAO Shushan, FAN Xiaohua

Journal of Xidian University. 2018, 45(6):  86-91.  doi:10.3969/j.issn.1001-2400.2018.06.015

Abstract ( 445 )   PDF (1592KB) ( 142 )   Save

References | Related Articles | Metrics

A frequency calibration technique with the hybrid calibration process is proposed to reduce the calibration time of the voltage-controlled oscillator (VCO) and maintain the fine calibration resolution. The calibration technique begins by employing the relative frequency comparison method to discern and decrease the difference between the frequency tuning curve and the target frequency. When the frequency tuning curve is close to the target frequency, the calibration technique changes to utilize the frequency error detection method to achieve a fine calibration resolution. With the fast relative frequency comparison process and the high-accuracy frequency error detection process, the calibration time for a frequency resolution of 2MHz is only 2.77μs. Compared to the conventional binary searcher process by only employing the frequency error detection method, the proposed technique saves 31% of the calibration time under the same calibration resolution. Furthermore, this calibration technique is suitable for the wideband fractional-N frequency synthesizer.

Spatial sensing matrix learning for PolSAR image classification

SUN Chen, CHENG Liye

Journal of Xidian University. 2018, 45(6):  92-98.  doi:10.3969/j.issn.1001-2400.2018.06.016

Abstract ( 342 )   PDF (1192KB) ( 114 )   Save

References | Related Articles | Metrics

To deal with the large scale problem, a classifier based on the spatial sensing matrix and Least Squares Support Vector Machine (LS-SVM) is proposed for the polarimetric synthetic aperture radar (PolSAR) image. Inspired by the compressive sensing theory, a spatial sensing matrix is designed, which is equal to the product of the measurement matrix and the kernel matrix. The discriminative sensing matrix is proposed to largely reduce the scale of the optimization problem. Then, by taking the special properties of the polarimetric data and spatial information into account, we propose a spatial-Wishart dictionary to reduce the noise of speckle. Finally, the compressive sensing inspired classifier is constructed and the sparse support vector coefficients are achieved. Classification accuracy and spatial consistency of the proposed classifier is superior to those of other classifiers.

Precoding superimposed training design with precoding gain optimization

WANG Qingbo, DOU Gaoqi, GAO Jun

Journal of Xidian University. 2018, 45(6):  99-105.  doi:10.3969/j.issn.1001-2400.2018.06.017

Abstract ( 269 )   PDF (623KB) ( 35 )   Save

References | Related Articles | Metrics

To gain a higher precoding gain of the precoder for the full-rank precoding superimposed training model, the diversity and coding gain of the precoder are derived by pair-wise error probability analysis, with optimization strategies provided. Furthermore, chirp sequence-based full-rank precoding superimposed training is proposed. The proposed scheme adopts chirp sequences to construct the precoder and training sequences, so as to realize orthogonal transmission for data and training sequences and achieve a greater coding gain than the existing scheme with the same performance of channel estimation. Simulation results show that the proposed scheme remarkably improves the performance of symbol detection in high signal noise ratio environment. The precoding superimposed training model constructed by chirp sequences achieves a greater coding gain than that constructed by polyphase sequences.

Novel sea ice edge detection method in SAR imagery

LIU Jiange, MU Dejun

Journal of Xidian University. 2018, 45(6):  106-111.  doi:10.3969/j.issn.1001-2400.2018.06.018

Abstract ( 368 )   PDF (4064KB) ( 72 )   Save

References | Related Articles | Metrics

The paper proposes an ice edge detection method in synthetic aperture radar (SAR) images especially where sea ice concentration is low and might be underestimated by using passive microwave data. According to the multi scales and orientations of curve-like features in the marginal ice zone (MIZ) in SAR images, the edge detection method is chosen to unify the curve features of different backscatters and contrasts into edge information. Based on the difference of edge information richness between MIZ and open water, the curvelet-based multiscale strategy active contour is used to detect the ice edge. Experimental results show a significant increase in accuracy compared with the ice edge definition from passive microwave sea ice concentration.

Interference neutralization transmission of two-cluster multi-user MIMO relay interference channels

ZHANG Bowei, WU Weinong, XIAO Jingwei, HUANG Jie, WANG Jian, FENG Wenjiang

Journal of Xidian University. 2018, 45(6):  112-117.  doi:10.3969/j.issn.1001-2400.2018.06.019

Abstract ( 287 )   PDF (649KB) ( 47 )   Save

References | Related Articles | Metrics

Based on the design of the relay beamforming matrix, an interference neutralization (IN) scheme is proposed for the two-cluster multi-user multiple-input multiple-output (MIMO) relay interference channel. First, the received signal at each relay is reconstructed by using the relay receive beamforming matrix. Then, each relay constructs the transmit beamforming matrix to eliminate the inter-cluster interference. The achievable degrees of freedom (DoF), the achievable sum rate and the application conditions for the IN scheme are analyzed. Experimental results have shown that the IN scheme can provide a significant DoF gain over the time division multiple access (TDMA) and the successive network coding (SNC) schemes.

Design of modified SCR devices and optimization of the leakage characteristics under ESD stress

LIU Huyun, LIANG Hailian, GU Xiaofeng, MA Yike, WANG Xin

Journal of Xidian University. 2018, 45(6):  118-122.  doi:10.3969/j.issn.1001-2400.2018.06.020

Abstract ( 370 )   PDF (1872KB) ( 49 )   Save

References | Related Articles | Metrics

In order to investigate the easy latch-up and soft failure problem of on-chip integrated circuit electrostatic discharge (ESD) protection, a silicon controlled rectifier (SCR) embedded with the N-bridge and MOS (SCR-N-MOS) is proposed. The transmission line pulse test results show that the voltage snapback margin of the SCR-N-MOS is reduced by about 28.6% compared to the conventional N-bridge modified SCR. However, the leakage current of the SCR-N-MOS degrades gradually from 2.8×10^-7A to 1.7×10^-5A when the ESD transient current increases from 2.0A to 3.2A, resulting in soft failure. The technology computer aided design simulations of the SCR-N-MOS indicate that the lattice temperature increases  up to 1160.5K under an ESD current stress of 10^-4A. By optimizing the layout of the SCR-N-MOS and the metal wiring, the power density crowding effect is weakened and the leakage current can be decreased and maintained in the order of 10^-9A under the same stress. Therefore, the layout optimization method can effectively suppress the local overheating and improve the thermal stability of the on-chip integrated circuit ESD protection.

Algorithm for near-field source localization using the minimum redundant symmetric array

XIAO Song, CHEN Hui, NI Mengyu, NI Liuliu, ZHANG Jiajia

Journal of Xidian University. 2018, 45(6):  123-129.  doi:10.3969/j.issn.1001-2400.2018.06.021

Abstract ( 291 )   PDF (622KB) ( 148 )   Save

References | Related Articles | Metrics

For the issue of aperture and array elements losses in near-field source localization when having a limited number of sensors, a covariance matrix reconstruction method is proposed based on the symmetric minimum redundant array which allows the element spacing not to be limited to the quarter wavelength. A fourth-order cumulant matrix is constructed which is related only to the azimuth to estimate the azimuth angles of sources by the multiple signal classification (MUSIC) algorithm. Then the ranges of sources can be estimated by searching the spectral peak according to the estimated angles. The algorithm extends the aperture of the array and enhances the degree of freedom. Simulation results show that the algorithm can estimate a  much larger number of sources and has a better estimated performance and spatial resolution. Besides, the algorithm only requires one-dimensional search, which avoids two-dimensional parameter matching.

Improved single epoch ambiguity resolution algorithm based on the dual frequency correlation method

QUAN Yuan, ZHAO Xiubin, PANG Chunlei, ZHANG Hao, WANG Yong, WU Shaoshi

Journal of Xidian University. 2018, 45(6):  130-136.  doi:10.3969/j.issn.1001-2400.2018.06.022

Abstract ( 303 )   PDF (754KB) ( 48 )   Save

References | Related Articles | Metrics

The dual frequency correlation method which is based on the relevant characteristic of dual frequency carrier phase measurement constructs the ambiguity error band to fix integer ambiguity quickly. However, the algorithm needs to set the pseudo-range measurement accuracy in advance, which results in a false ambiguity search scope or an ineffective filter efficiency easily. First, the information on the baseline length and the gaze vector is utilized to derive the double difference geometric correlation model in order to obtain the integer ambiguity search range quantitatively. Then, the relationship among the sparsity of integer ambiguity search space, carrier phase wavelength, length and bandwidth of the error band are analyzed. Finally, the search range of the wide lane ambiguity is regarded as the horizontal axis of the error band axis to strengthen the sparsity of the search space. Consequently, an improved single epoch ambiguity resolution algorithm based on the dual frequency correlation method is proposed. Compared with the original algorithm in the single epoch, experimental results show that the average constringent rate of single-dimensional ambiguity increases to 85.71%, that the average computational cost is reduced to 0.46%, and that the success rate is increased from 95.28% to 99.78%, which greatly improves the reliability of ambiguity solution.

High precision common mode charge error fore-ground calibration circuit for the charge-domain ADC

CHEN Zhenhai, WEI Jinghe, YU Zongguang, SU Xiaobo, XUE Yan, ZHANG Hong

Journal of Xidian University. 2018, 45(6):  137-143+149.  doi:10.3969/j.issn.1001-2400.2018.06.023

Abstract ( 386 )   PDF (2092KB) ( 93 )   Save

References | Related Articles | Metrics

A mix-signal high precision common mode charge error calibration circuit is proposed. The calibration circuit can be used to compensate the common mode charge errors caused by the deviation of the charge transfer cutoff voltage of boosted charge transfer introduced by PVT variation, the variation of input common mode charge and the capacitor mismatch in the pipelined sub-stage circuit in the charge domain pipelined ADCs. Based on the proposed calibration circuit, a 14bit 210MS/s charge domain pipelined ADC is designed and realized in a 1P6M 0.18μm CMOS process. Test results show that the 14bit 210MS/s ADC achieves the signal-to-noise ratio of 71.5dBFS and the spurious free dynamic range of 85.4dB, with 30.1MHz input single tone signal at 210MS/s, while the ADC core consumes the power of 205mW and occupies an area of  3.2mm².

Face recognition method for integrating the nested residual CNN and angular metric

LIANG Xiaoxi, CAI Xiaodong, WANG Meng, KU Haohua

Journal of Xidian University. 2018, 45(6):  144-149.  doi:10.3969/j.issn.1001-2400.2018.06.024

Abstract ( 326 )   PDF (557KB) ( 57 )   Save

References | Related Articles | Metrics

Softmax loss is usually used in convolution neural networks for feature learning, but features obtained are not discriminative enough in some cases. Traditional methods for trying to solve this problem come with additional computational complexity. A face recognition method for integrating the nested residual convolution neural network and angle metric is proposed. First, a novel feature extraction network based on the nested residual block is designed to extract various features by integrating feature maps. Then, a method of angular metric based on weight normalization is utilized. The discrimination of features is enhanced by normalizing the weights of the last fully connected layer. The learned features can satisfy the condition that the maximum intra-class distance is less than the minimum inter-class distance by combining two methods mentioned above for training. Experimental results indicate that this method leads to an accuracy of 99.03% on the LFW(Labeled Faces in the Wild). The proposed algorithm only contains a single network and provides a higher accuracy and a lower computational cost than those methods using softmax loss and other metric learning.

Automatic assessment of children's vision

LI Xue, AI Lirong, ZHOU Xiaojing, ZHANG Kai

Journal of Xidian University. 2018, 45(6):  150-155.  doi:10.3969/j.issn.1001-2400.2018.06.025

Abstract ( 334 )   PDF (1629KB) ( 109 )   Save

References | Related Articles | Metrics

Aiming at the difficulty of children's vision examination, an automatic evaluation method for children's vision examination video is proposed based on the synchronization system of children's vision examination. In this method, the target detection algorithm is used to locate important objects such as the vision card and the examinee's eyes in the video, the doctor's experience is simulated to analyze the child's gaze situation on different grades of vision cards, and then an image classification algorithm is used for the bars on the critical vision card. Hence the visual acuity level is evaluated by the above information. The accuracy of both key object detection and vision card classification has reached more than 90%, and the accuracy of automatic vision assessment for children has reached 70%. Therefore the method of automatic assessment for children's vision level is convenient and objective, which is of important reference value for clinicians.

Lane-level path guiding method for an unmanned ground vehicle

YANG Qiangrong, WANG Meiling, YU Huachao

Journal of Xidian University. 2018, 45(6):  156-161.  doi:10.3969/j.issn.1001-2400.2018.06.026

Abstract ( 371 )   PDF (4187KB) ( 69 )   Save

References | Related Articles | Metrics

A lane-level digital map is constructed with multi-sensors to meet the requirements of an unmanned ground vehicle for a high-precision digital map and the needs to navigate in the structured road environment. Based on this lane-level digital map, a lane-level path guiding algorithm is also proposed. First, a lane-level digital map is constructed with the multi-sensor data collected by the data collecting platform and the lanes and segments are built according to traffic rules. Next, different property tables are designed for different geographic elements to represent their characteristics. Then, a global path planning method is proposed according to the characteristic of the structured road environment. Finally, a local path planning method is proposed to avoid the obstacles when the UGV meets the obstacles. According to the results of both on-vehicle experiments and contests for unmanned vehicles, it is shown that the proposed digital map and path guiding method meet the requirements of unmanned ground vehicles in the structured road environment and offer a reliable guidance and guarantee for fast and safe driving of unmanned ground vehicles.

Efficient pseudo-codeword search algorithm for LDPC codes

GUO Junjun, BAI Shuodong, MU Jianjun, JING Xin, XIAO Feng

Journal of Xidian University. 2018, 45(6):  162-166.  doi:10.3969/j.issn.1001-2400.2018.06.027

Abstract ( 295 )   PDF (514KB) ( 67 )   Save

References | Related Articles | Metrics

In order to obtain the pseudo-codewords which can decrease the linear programming (LP) decoding performance of low-density parity-check(LDPC) codes, an efficient algorithm for searching low-weight pseudo-codewords of LDPC codes is proposed on the binary symmetric channel by making a detailed analysis of the corresponding relationship between variable nodes in the harmful Tanner subgraphs and input components of LP decoders. The proposed algorithm can rapidly converge to low-weight pseudo-codewords of LDPC codes with finite iterations by biasing noises for the input vectors of the LP decoder based on the alternating direction method of multipliers. Simulation results show that the proposed approach can accurately find many low-weight pseudo-codewords of regular and irregular LDPC codes with a short or middle length compared with the existing pseudo-codeword search method.