2018 No. 10-Application of Electronic Technique

Special Column-Artificial Intelligence

Handwriting digital recognition and application based on TensorFlow deep learning

DOI：10.16157/j.issn.0258-7998.182249

Author：Huang Rui，Lu Xuming，Wu Yilin

Author Affilications：Department of Computer Science，Guangdong University of Education，Guangzhou 510303，China

Abstract：The recognition of handwritten digits is an important part of the artificial intelligence recognition system. Due to the difference in individual handwritten numbers, the existing recognition system has a lower accuracy rate. This paper is based on the TensorFlow deep learning framework to complete the recognition and application of handwritten numbers. Firstly, the Softmax and Convolutional Neural Network(CNN) model structure is established and analyzed. Secondly, deep learning is performed on 60 000 samples of the handwritten data set MNIST, and then 10 000 samples are tested and compared. Finally, the optimal model is transplanted to the Android platform for application. Compared with the traditional Softmax model, the recognition rate based on TensorFlow deep learning CNN model is as high as 99.17%, an increase of 7.6%, which provides certain scientific research value for the development of artificial intelligence recognition system.

Key word：
TensorFlow

deep learning

Convolutional Neural Network(CNN)

digital recognition

View(17044)

Human behavior recognition algorithm based on deep learning

DOI：10.16157/j.issn.0258-7998.182201

Author：Huang Youwen，Wan Chaolun

Author Affilications：School of Information Engineering，Jiangxi University of Science and Technology，Ganzhou 341000，China

Abstract：In order to improve the problem of low accuracy in human behavior recognition task, a neural network based on batch normalization convolution neural network(CNN) and long short-term memory(LSTM) neural network is proposed. The CNN part introduces the idea of batch normalization, and the training data of the input network are normalized in mini-batch. After full connection, they are sent to long short-term memory neural network. The algorithm adopts the space-time dual stream network model structure. The RGB image of video data is taken as spatial stream network input, and the optical flow field image is taken as time flow network input. Then the recognition results obtained by the time-space dual-stream network are combined in a certain proportion to obtain the final behavior recognition result. The experimental results show that the space-time dual stream neural network algorithm designed in this paper has a high recognition accuracy in human behavior recognition tasks.

Key word：
behavior recognition

batch normalization

long short-term memory neural network

deep learning

View(19166)

Research on finger vein recognition based on capsule network

DOI：10.16157/j.issn.0258-7998.182236

Author：Yu Chengbo，Xiong Dien

Author Affilications：School of Electrical and Electronic Engineering，Chongqing University of Techology，Chongqing 400050，China

Abstract：This paper propose a finger vein recognition algorithm based on the CapsNets(Capsule Network for short) to solve the problem of the information loss of the finger vein in the Convolution Neural Network(CNN). The CapsNets is transferred from the bottom to the high level in the form of capsule in the whole learning process, so that the multidimensional characteristics of the finger vein are encapsulated in the form of vector, and the features will be preserved in the network, but not in the network after the loss is recovered. In this paper, 60 000 images are used as training set, and 10 000 images are used as test set. The experimental results show that the network structure features of CapsNets are more obvious than that of CNN, the accuracy of VGG is increased by 13.6%, and the value of loss converges to 0.01.

Key word：
CapsNets

finger vein recognition

deep learning

CNN

View(11958)

Design of real-time recognition hardware system framework based on deep learning

DOI：10.16157/j.issn.0258-7998.182417

Author：Wang Kun，Zhou Hua

Author Affilications：College of Big Data and Information Engineering，Guizhou University，Guiyang 550025，China

Abstract：This paper designs a real-time recognition hardware system framework based on deep learning. The system framework uses Keras to complete the training of the convolutional neural network model and extracts the parameters of the network. Using the FPGA+ARM software and hardware coordination method of the ZYNQ device, ARM was used to complete the acquisition, preprocessing and display of real-time image data. Through the FPGA，the hardening of the convolutional neural network is performed and the image is recognized, and the recognition result is sent to the upper computer for real-time display. The system framework uses MNIST and Fashion MNIST data sets as network model hardening test samples. The experimental results show that the system framework can display and identify image data in real time and accurately under the general scene. And it has the characteristics of high portability, fast processing speed and low power consumption.

Key word：
deep learning

CNN

ZYNQ

hardware-software-collaborative

View(11762)

Image coloring based on convolution neural network

DOI：10.16157/j.issn.0258-7998.182404

Author：Xu Zhonghui，Lv Weishuai

Author Affilications：Faculty of Information Engineering，Jiangxi University of Science and Technology，Ganzhou 341000，China

Abstract：The goal of image coloring is to assign color to each pixel of the grayscale image, which is a hot topic in the field of image processing. U-Net is used as the main line network, and a fully automatic coloring network model is designed based on deep learning and convolutional neural networks. In this model, the branch line uses the convolutional neural network SE-Inception-ResNet-v2 as a high-level feature extractor to extract the global information of the image, and the Power Linear Unit(PoLU) function is used to replace the Rectified Linear Unit(ReLU) function in the network. The Experimental results show that this coloring network model can effectively color grayscale images.

Key word：
colorization

convolutional neural network

deep learning

Squeeze-and-Excitation Networks(SENet) model

Power Linear Unit(PoLU)

View(10954)

Review and Comment

Overview of energy harvesting cognitive radio sensor networks

DOI：10.16157/j.issn.0258-7998.180940

Author：Jiao Wanguo，Li Yurong，Zhou Wen

Author Affilications：School of Information Science and Technology，Nanjing Forestry University，Nanjing 210037，China

Abstract：In energy harvesting cognitive radio sensor networks(EHCRSNs), energy harvesting supplies continual energy to eliminate the limitation of limit-battery sensor, while cognitive radio mitigates the scarcity of spectrum resource. Hence, the operation of EHCRSNs can be continually and effectively executed. However, due to the dynamic and stochastic property of the energy harvesting progress and the available spectrum, there are some significant challenges in the area of network resource management of EHCRSNs. This paper will firstly introduce the background of EHCRSNs. Then, the research on EHCRSNs is systematically summarized and discussed. At last, we point out limitations of these studies and possible future research directions.

Key word：
wireless sensor networks

energy harvesting

cognitive radio

resource management

View(9554)

Microelectronic Technology

Etching technique of high aspect ratio silicon trenches based on CMOS-MEMS process

DOI：10.16157/j.issn.0258-7998.180945

Author：Zhang Haihua1，Lv Yufei2，Lu Zhongxuan1

Author Affilications：1.Beijing Institute of Spacecraft System Engineering，Beijing 100094，China； 2.Beijing Space era Laser Navigation Technology Co.，Ltd.，Beijing 100094，China

Abstract：Etching technique of deep silicon trenches with high aspect ratio is critical in improving integration and accuracy of Micro-Electro-Mechanical Systems(MEMS)sensor array and reducing the cost of it. To fabricate silicon isolation trenches with aspect ratio equaling to 25 μm/0.8 μm,we use and compare three kinds of MEMS process，Reactive Ion Etching(RIE),Bosch,and RIE combined with Bosch,which are all CMOS process compatible and easy to integrate with CMOS ICs.Finally it proves that RIE combined with Bosch process is better than the other two in producing deep trenches with smooth side walls.

Key word：
CMOS-MEMS process

high aspect ratio

deep trench etching

RIE combined with Bosch

View(13814)

Realization of SRRC filter and multi-rate conversion based on FPGA

DOI：10.16157/j.issn.0258-7998.180491

Author：Yang Yang1，Yan Zheng2，Liu Minwei2，Dong Jicheng2

Author Affilications：1.Integrated Circuit Design Center，Aerors Incorporated Company Limited.Sichuan Branch，Chengdu 610100，China； 2.Integrated Circuit Design Center，Space Star Technology CO.，LTD.，Beijing 100086，China

Abstract：In order to eliminate the inter symbol interference(ISI) in the communication system and improve the band utilization, the square root-raised cosine filter is often used to realize the base-band shaping filter of the baseband signal. In order to achieve high rate transmission of signals with different symbol rates in communication systems, multi-rate conversion techniques are used in digital signal processing to improve the sampling rate of digital signals. In this paper, SRRC filter and HB, CIC and Farrow filters are cascaded, then a multi-rate converter module based on FPGA is realized. The module can achieve arbitrary multiples of upper sampling , what’s more, by online coefficient reload capability of SRRC filter and online upper insertion ratio reload capability of CIC and Farrow filter, it saves FPGA internal resources effictively. SRRC filter and multi-rate converter module is realized through Verilog programming and the use of IP core on ISE platform. The simulation waveforms of ModelSim and the experimental results are given, raised cosine and multi-rate conversion characteristics are verified, what’s more, the multi-rate conversion module can achieve arbitrary multiples of upper sampling, eliminate the inter symbol interference and improve the band utilization. It is simple to implement effectively.

Key word：
FPGA

base-band shaping filter

upper sampling

arbitrary multiples

ISI

View(16504)

Implementation of wavelet transform based on switched current circuits

DOI：10.16157/j.issn.0258-7998.180443

Author：Hu Qinchun1，Huo Ping2，You Xiaoquan1，Zhang Song1，Guo Lifang1，Tan Huayong1

Author Affilications：1.School of Network and Communication Engineering，Chengdu Technological University，Chengdu 611730，China； 2.School of Electrical Engineering，Chengdu Technological University，Chengdu 611730，China

Abstract：In order to meet the engineering application requirement for low-power and real-time signal processing, an implementation of wavelet transform using switched current(SI) technique is presented in this paper. The working principles of wavelet transform, both in time-domain and frequency-domain are discussed. The circuits are composed of analog filter banks whose impulse responses are the first derivative of Gaussian. By adjusting the clock frequency of SI circuits, the wavelet transform in differential scales are able to be performed. The good performance of the proposed method is verified by simulated results.

Key word：
wavelet transform

time-domain

frequency-domain

switched current

View(9184)

Indirect alignment of MEMS inertial navigation system based on multi forgetting factors adaptive Kalman filter de-noising

Author：Sun Jia，Zou Jing，Hu Tong

Author Affilications：Shandong Provincial Key Laboratory of Ocean Environmental Monitoring Techno1ogy， Institute of Oceanographic Instrumentation，Qilu University of Technology(Shandong Academy of Sciences)，Qingdao 266001，China

Abstract：According to the problem that the low output precision of MEMS inertial devices leads to large initial alignment error of inertial navigation system，an indirect coarse alignment method for MEMS was proposed, in which multi forgetting factors adaptive Kalman filter(MFFAKF) was used at first to reduce the noise of original data and then the magnetometer was used to assist. The original data from the accelerometers and magnetometers were pretreated by using multiple forgetting factors to adjust the different filtering channels, and then the attitude angle was calculated. The initial alignment accuracy was improved and the alignment time was shorter by this method. The experimental results showed that the coarse alignment heading accuracy obtained by MFFAKF satisfied the coarse alignment requirements of the MEMS inertial navigation system.

Key word：
MEMS

inertial navigation system

forgetting factors adaptive

attitude calculation

View(11910)

The design of a DC-30 GHz GaAs pHEMT distributed power amplifier

DOI：10.16157/j.issn.0258-7998.180349

Author：Liu Yanpeng，Wei Qidi，Zhang Guohao

Author Affilications：School of Information Engineering，Guangdong University of Technology，Guangzhou 510006，China

Abstract：This paper describes a distributed power amplifier(DA) which was developed using 0.25 μm GaAs pHEMT process. The process of its circuit design and optimization is presented in details. Gain flatness in the low frequency range of this DA has a significant improvement by employing low frequency termination. The simulation result shows that DA which operates between DC and 30 GHz has a gain of 8.5 dB and 21 dBm of output power at 1 dB gain compression and maximum PAE up to 20%.

Key word：
distributed power amplifier

GaAs

pHEMT

View(11266)

Design of domino logic obfuscation circuit based on dummy contacts

DOI：10.16157/j.issn.0258-7998.180395

Author：Li Liwei，Wang Pengjun，Zhang Yuejun

Author Affilications：Institute of Circuits and Systems，Ningbo University，Ningbo 315211，China

Abstract：Through the study of the logic obfuscation circuit and the physical level camouflage technology, a scheme of the domino logic obfuscation circuit is proposed based on dummy contacts, which can prevent reverse engineering. Firstly, the scheme adopts the same circuit structure to realize the NAND, NOR, INV logic functions by contacts′ reality. Then, the TSMC 65 nm is used to compile the ISCAS-89 benchmark circuit and the domino logic obfuscation circuit is applied to the circuit netlist. Finally, the netlist performance test after replacement design is completed. Compared with the existing circuits, the power consumption of the designed circuit is reduced by 4.79%, the area overhead is reduced by 2.16%, and the delay time is reduced by 16.66%, which effectively improves the circuit′s safety.

Key word：
logic obfuscation

dummy contact

domino logic

hardware security

View(10362)

A high speed truly random number generator based on thermal noise oscilltor

DOI：10.16157/j.issn.0258-7998.180002

Author：Wei Zikui1，2，Fu Ling1，2，Wang Xue3，He Yang1，2，Jin Xin1，2，Tan Lang1，2， Hu Yi1，2，Tang Xiaoke1，2，Zhang Haifeng1，2，Zhao Dongyan1，2

Author Affilications：1.State Grid Key Laboratory of Power Industrial Chip Design and Analysis Technology， Beijing Smart-Chip Microelectronics Technology Co.，Ltd.，Beijing 100192，China； 2. Beijing Engineering Research Center of High-reliability IC with Power Industrial Grade， Beijing Smart-Chip Microelectronics Technology Co.，Ltd.，Beijing 100192，China； 3.Information and Communication Branch of Liaoning Electric Power Co.，Ltd.，Shenyang 110006，China

Abstract：A truly random number generator based on thermal noise was proposed, and the resistance thermal noise was used as the noise source of the low frequency clock. It uses two independent working relative relationship between high and low frequency oscillator to sample noise source, and uses low frequency oscillator with jitter to sample the high frequency osciuator which has a fixed cycle oscillator by D flip-flop, to generate a random number sequence. The circuit was based on SMIC 28 nm CMOS process. The simulation results show that based on the thermal noise of truly random number generator the output throughput rate range is 1.8~12 Mb/s, random number output satisfies the requirement of entropy source AIS31 truly random number test, and the overall power consumption is 0.388 mW.

Key word：
truly random number generator

thermal noise

oscillator

random tremble

power consumption

View(14542)

Embedded Technology

TRDP protocol stack transplantation on VxWorks

DOI：10.16157/j.issn.0258-7998.181196

Author：Zhou Haoshang1，2，Gao Zhaoling3

Author Affilications：1.CRRC Dalian R&D CO.，LTD.，Dalian 116052，China； 2.State Key Laboratory for Traction and Control System of EMU and Locomotive，Dalian 116052，China； 3.Department of Electronic Engineering，Dalian Neusoft University of Information，Dalian 116023，China

Abstract：Firstly,in the VxWorks operating system, the problems encountered in the train real-time data protocol(TRDP) stack transplant are analyzed and solved, and the solutions to the failure of data transmission in the process of TRDP application are encountered. Finally, the common backdoor problem in the current VxWorks 6.5 version FTP component are pointed out. After two years' verification, and assessment of the application of over 200 thousand kilometers,the problems encountered in the application of the protocol stack in the locomotive and vehicle network control system and the solution in the application of the protocol stack in the locomotive and vehicle network control system can prove that the transplant successes, the method is reliable, and it is worth popularizing.

Key word：
VxWorks 6.5

TRDP protocol stack

CCU

View(16554)

Design and implementation of fiber-optic vibration positioning system based on FPGA

DOI：10.16157/j.issn.0258-7998.181072

Author：Luo Yijun，Fang Li

Author Affilications：Electronic Information School，Wuhan University，Wuhan 430072，China

Abstract：Aiming at the problem that the perimeter of defense zone can’t be located in real time when it is invaded, this paper designed and realized a real-time distributed optical fiber vibration positioning system. The optical path was built on the basis of Mach-Zehnder fiber interference technique，FPGA combined with high speed ADC was used as hardware platform to transform and collect the two signals from the optical path,then the cross correlation was taken as the main algorithm to process the acquired signal to get the vibration position. Tested on a 160-meter fiber, the system can give the vibration position within 500 ms after the vibration occurred. Seting the sampling rate to 10 MHz and after multiple positioning test, the actual positioning error of the system is ±10 m. And the positioning does not need the upper computer, which improves the device portability and reduces the cost.

Key word：
FPGA

optical fiber vibration

cross-correlation

positioning

View(18726)

Pulse generator with real-time industrial ethernet

DOI：10.16157/j.issn.0258-7998.180998

Author：Lin Hao，Han Qingmin，Song Dong，Chen Hai

Author Affilications：The 6th Research Institute of China Electronics Corporation，Beijing 100083，China

Abstract：The pulse generator is the basic equipment for device drivers and control, widely used in many occasions, such as power electronics, motion control, vehicle control, etc. However, the used pulse generator is weak in communication, which cannot meet high-speed control requirements or be used in the intelligent manufacturing application. To solve the above problem, this paper designed a pulse generator with real-time industrial ethernet EtherCAT and Loongson processor LS1C. The designed pulse generators can form the daisy chain and ring topology by EtherCAT. And the pulse generator is very flexible and suited for many different applications with its four operation modes. Finally, experiment results are presented to demonstrate the operation modes and performance.

Key word：
real-time industrial ethernet

EtherCAT

pulse generator

View(12248)

Design of automatic irrigation system after bladder surgery

DOI：10.16157/j.issn.0258-7998.180950

Author：Chen Hongchao，Ye Bing，Bian Shiya

Author Affilications：School of Electronic Science and and Applied Physics，Hefei University of Technology，Hefei 230009，China

Abstract：According to the development trend of nursing care based on urinary tract surgery，an automatic irrigation system after bladder operation was designed. The system uses OV2640 image sensor to capture the image of the patient′s drainage fluid, the STM32 microcontroller receives the captured image and displays it on the LCD screen. In the program, the image frame of the drainage fluid was analyzed by mean gray value and erosion algorithm, and it judged the concentration of blood dissolved in the drainage fluid and whether it is mixed with clot. This method has the advantages of small calculation and fast processing. After the test experiment, the results show that the system achieves automatic detection of blood and blood clot dissolved in the drainage fluid, and ensures the efficiency and accuracy．

Key word：
STM32 microcontroller

OV2640 image sensor

mean gray value

erosion algorithm

automatic detection

View(7932)

A pavement damage location algorithm based on machine vision distance compensation

DOI：10.16157/j.issn.0258-7998.181149

Author：Miao Chengyu1，2，3，Li Peng1，2，3，Zhang Jingrong1，2，3，Gu Binbin1，2，3，Jiang Wei1，2，3

Author Affilications：1.Jiangsu Collaborative Innovation Center on Atmospheric Environment and Equipment Technology， Nanjing University of Information Science and Technology，Nanjing 210044，China； 2.Jiangsu Meteorological Sensor Network Technology Engineering Center，Nanjing University of Information Science and Technology， Nanjing 210044，China； 3.Jiangsu Key Laboratory of Meteorological Observation and Information Procession， Nanjing University of Information Science and Technology，Nanjing 210044，China

Abstract：The real-time and accurate automatic detection technology of road surface damage has important significance for highway management and maintenance. This paper combines the FPGA-based pavement damage detection system to research the positioning accuracy. Due to the influence of factors such as the installation perspective of the system equipment and the speed of the carrier, the accuracy of the road surface positioning may be reduced. This paper proposes a pavement damage location algorithm based on machine vision distance compensation. The experimental results show that when the vehicle′s driving speed is below 72 km/h, the positioning error is reduced from 5 m to 2 m. It meets the requirements of the system design.

Key word：
pavement damage

automatic detection

extended Kalman filtering

machine vision

distance compensation

View(8818)

Measurement Control Technology and Instruments

Development of portable orthogonal lock-in amplifier based on FPGA

DOI：10.16157/j.issn.0258-7998.181096

Author：Xie Guihui1，2，Zheng Xuchu3，Zhao Tianming3，Liu Zixu3，Zhao Juan1，2

Author Affilications：1.School of Automation, China University of Geosciences，Wuhan 430074，China； 2.Hubei Key Laboratory of Advanced Control and Intelligent Automation for Complex Systems，Wuhan 430074，China； 3.School of Mechanical and Electronic Information，China University of Geosciences，Wuhan 430074，China

Abstract：A portable orthogonal phase-locked amplifier for weak signal detection has been developed based on FPGA. First,the signal processing module receives the signal,does variable gain amplification on the signal and puts it through the power frequency noise filtering. The signal is put into the FPGA through A/D conversion module conversion. By the digital phase-locked loop ,the amplifier completes the measured signal phase lock, extracts the feedback signal to calculate the measured signal amplitude, and completes weak signal detection. The lock-in amplifier amplitude measurement range is 100 nV~100 μV, dynamic range can reach up to 60 dB, phase accuracy can attain 0.001°. The system is designed as a portable device, low cost, easy to maintain, can meet the measurement requirements of the special environment such as fieldwork or etc, and has broad application prospects.

Key word：
FPGA

weak signal detection

portable

phase lock zoom

View(28064)

Denoising method of ESD current waveform based on wavelet and adaptive Kalman filtering

DOI：10.16157/j.issn.0258-7998.180921

Author：Zhou Kui1，2，Ruan Fangming2，3，Guan Sheng1，Su Ming3，Wang Heng3

Author Affilications：1.School of Big Data and Information Engineering, Guizhou University，Guiyang 550025，China； 2.Beijing Oriental Institute of Measurement and Test，Beijing 100094，China； 3.School of Big Data and Computer Sciences，Guizhou Normal University，Guiyang 550001，China

Abstract：In order to reduce the interference noise of ESD current signal, wavelet analysis and adaptive Kalman filter are used in the study of ESD current waveform denoising. The numerical solutions of ESD current of human body-metal model are computed by the Adams prediction-calibration algorithm，and a corresponding noisy ESD current signal model is established. The wavelet denoising method performs denoising performance analysis on this model. Aiming at the measured ESD current waveform, the wavelet threshold denoising method is used to preprocess the ESD current waveforms, and obtain more stable observation data. According to the information of the observed data, the adaptive Kalman filter algorithm proposed by Sage-Husa is used to optimize the data processed the wavelet pretreatment data. The results show that the wavelet analysis and adaptive Kalman filter algorithm can effectively reduce the interference noise of the measured ESD current waveform.

Key word：
ESD

current waveform

Adams prediction-correction algorithm

wavelet denoising

adaptive Kalman filter

View(9902)

Hybrid optimization algorithm of improved artificial potential field and TAS-RRT

DOI：10.16157/j.issn.0258-7998.181374

Author：Xu Xiaohui1，Zhang Jinlong2

Author Affilications：1.School of Architectural Engineering，The City Vocational College of Jiangsu，Nanjing 210000，China； 2.School of Electrical and Automation Engineering，Nanjing Normal University，Nanjing 210042，China

Abstract：To solve the problem of artificial potential field algorithm(APF) being liable to fall into local minimum，this paper presents a strategy based on angle of rotating speed vectors to accurately locate the position of jump point.Combined with improved APF,the RRT based on transition of angle different of rotating speed vectors(TAS-RRT)can be used to path planning dynamicly. Firstly, artificial potential field algorithm is used in obstacle avoidance motion planning. When falling into local minimum,the sampling strategy of the basic RRT is improved to adaptively seek the target point and the connection mode of the local planner is adjusted to change exploration rate of the tree,so that the search process can be made to jump out of the attractive areas of local minimum point quickly.In addition, APF will be applied again when the the sampling point meets the stop condition of angle of rotating speed vectors. The simulation experiments show that the control precision and velocity of the motion planning are enhanced with TAS-RRT guiding the sampling nodes to the target one gradually and quickly.Besides that,the method can be applied to motion planning of different obstacle environment.

Key word：
motion planning of robot

artificial potential field

angle of rotating speed vectors

TAS-RRT

View(16068)

Design and experimental study of MEMS personal navigation system

DOI：10.16157/j.issn.0258-7998.181128

Author：Qian Wengao1，Chen Jingjie1，Ma Hongyan2

Author Affilications：1.College of Electronic Information and Automation，Civil Aviation University of China，Tianjin 300300，China； 2.Basic Experiment Center，Civil Aviation University of China，Tianjin 300300，China

Abstract：Micro-Electro-Mechanical System(MEMS) personal inertial navigation system is designed. Based on the traditional inertial navigation algorithm, the system introduces zero speed correction technology and combines the characteristics of personal walking. First, the four elements method is used to construct the observation equation，and the time sampling increment method of the Bi card algorithm is used to solve the equation. Then the attitude of the inertial navigation device is obtained by extended Kalman filter algorithm. Zero speed detection algorithm and Kalman decomposition algorithm based on UD decomposition are used to correct velocity and position errors. Finally, the effectiveness of the algorithm is verified by indoor walking test, which lays the foundation for further research on improving indoor navigation accuracy.

Key word：
personal navigation

MEMS

extended Kalman Filter

zero-velocity detection and correction

View(10344)

Research on indoor positioning algorithm based on improved fingerprint library accuracy

DOI：10.16157/j.issn.0258-7998.180768

Author：Wang Peiliang1，2，Zhang Ting3，Xiao Yingjie1

Author Affilications：1.Engineering Research Center of Simulation Technology of the Ministry of Education， Merchant Marine College，Shanghai Maritime University，Shanghai 201306，China； 2.Weifang University of Science and Technology，Weifang 262700，China； 3.Shandong Transport Vocational College，Weifang 261206，China

Abstract：With the continuous development of sensor networks, location-based services have gradually become the research hotspots, and indoor positioning technology is one of the most which are the fastest developing technology. In order to accurately and quickly position the location of indoor nodes to be measured, this research based on ZigBee wireless network proposes to use the algorithm based on outlier detection and double threshold filtering (frequency threshold and mean threshold) to process the RSSI data ,which are from different access points(APs), to establishes a high-precision fingerprint database. Then the final position of the node to be measured is calculated by using fuzzy C-Means(FCM) and the weighted K-nearest neighbor(WKNN) algorithm based on the frequency factor, and the“spare location” is also obtained. The result shows that the fingerprint database established by this research method has a great improvement in accuracy and the positioning accuracy has also been significantly improved.

Key word：
indoor positioning

ZigBee wireless network

fuzzy c-means

weighted k-nearest neighbor

View(14306)

A RFID and WSN based distributed localization algorithm

DOI：10.16157/j.issn.0258-7998.181472

Author：Lv Xiaowei

Author Affilications：Information Science Academy of China Electronics Technology Group Corporation，Beijing 100086，China

Abstract：Radio Frequency Identification（RFID）and Wireless Sensor Network（WSN）are both widely adopted in the application of target recognition/tracking and mobile positioning. In view of the technical limitation in the field of positioning application, such as the restrain of sense range, recognition efficiency, networking capability, and so on，this paper proposes a distributed localization algorithm, combining the advantages of RFID and WSN technologies with better object identification, communication and sensing capabilities. This method firstly employs RF reader and tags to obtain an accurate reference zone, and then uses range-free fingerprinting based localization method to make estimates in selected zone. Simulation results reveals that the proposed method improves effectively the localization accuracy of mobile nodes, as well as saves network resources.

Key word：
RFID

WSN

mobile positioning

distributed WSN

fingerprinting

View(11126)

The design and implementation of training system based on virtual reality technology

DOI：10.16157/j.issn.0258-7998.182387

Author：Liu Jin，Zhang Pengwang

Author Affilications：Dali Bureau，CSG EHV Transmission Company，Dali 671000，China

Abstract：With the development of computer technology and virtual technology, virtual reality technology has been widely applied to training, learning and simulation in many fields, such as architecture, medicine, electric power, aerospace, military，etc. By using virtual reality technology, people can simulate a limitless virtual space in a limited real space and people can also restore many dangerous operations in an absolutely safety condition. In the various environments simulated by virtual reality technology, the experiencer can obtain immersive interactive experience, and can keep repeating training, and can study in-depth. This article mainly introduces the design and implementation of the virtual reality training system which has been promoted and applied in China Southern Power Grid Company Limited and has got a good effect.

Key word：
virtual reality technology

training

immersive

interactive

View(13678)

Communication and Network

An anti-collision algorithm based on continuous slot detection

DOI：10.16157/j.issn.0258-7998.180393

Author：Yang Fan1，Ren Shougang2，Hao Shuixia1，Zhou Jun3，Yuan Peisen2

Author Affilications：1.School of Mathematics and Statistics，Jiangsu Normal University，Xuzhou 221116，China； 2.School of Information Sciences and Technology，Nanjing Agricultural University，Nanjing 210095，China； 3.Jiangsu Key Laboratory for Intelligent Agricultural Equipments，Nanjing Agricultural University，Nanjing 210031，China

Abstract：In order to further increase the speed of tag identification, for solving the insensitivity of the frame adjustment in dynamic frame slot aloha algorithm and the high computing complexity of the Q value adjustment in Q algorithm, a new anti-collision algorithm, called continuous slot detection(CSD), is proposed. The idea of algorithm and the process of operation are introduced. In this algorithm, the optimal efficient is obtained by detecting the status of consecutive slots in frame and setting frame length by rules. The simulation results show that compared with other anti-collision algorithm, the proposed algorithm can decrease identification delay by 4% and increase identification speed by 6%.

Key word：
RFID

anti-collision algorithm

Q value

slot detection

View(9456)

Poetry steganalysis based on convolutional neural network

DOI：10.16157/j.issn.0258-7998.180439

Author：Jin Peng1，2，Yang Zhongliang1，2，Huang Yongfeng1，2

Author Affilications：1.Department of Electronic Engineering，Tsinghua University，Beijing 100084，China； 2.Tsinghua National Laboratory of Information Science and Technology，Beijing 100084，China

Abstract：Using Chinese poetry as a carrier to realize text information hiding(IH) is a very hot research topic in the resent years. This kind of method has the advantages of large capacity and high concealment. However there isn′t a very effective steganalysis detection algorithm according to this kind of method. This paper puts forward a kind of poetry steganalysis detection algorithm based on convolution neural network(CNN),which can automatically judge whether the text of the input contains hidden information or not. Through training and testing the model by a large amount of data, the experiment results show that the proposed model can achieve 94.8% of detection accuracy, which shows very high poetry steganographic detection performance.

Key word：
convolutional neural network

information hiding

steganography detection

View(11902)

A replica algorithm based on popularity for P2P streaming media

DOI：10.16157/j.issn.0258-7998.180559

Author：Yang Ge1，2，Gao Bing1，2，Huang Jing1，He Hui1

Author Affilications：1.College of Information Technology，Beijing Normal University(Zhuhai Campus)，Zhuhai 519087，China； 2.Engineering Lab on Intelligent Perception for Internet of Things(ELIP)，Shenzhen Graduate School， Peking University，Shenzhen 518055，China

Abstract：In this paper，a new formula of popularity was proposed. It included the term deficit bandwidth and was based on the new popularity. Those streaming media files which need to be replicated were determined. A concept named comprehensive performance indicators was proposed. Peers were sorted by their comprehensive performance indicators and those peers with high comprehensive performance indicators had priority to place these popular files. Replacement algorithm would be carried out if there was no enough space to cache the new file, and the file which had largest ratio of its duplicates to the desired duplicates would be replaced by the new file. Experimental results show that the workload of the proposed algorithm, compared with proportional replication algorithm, is in steady conditions earlier about 13% in advance and is smaller. Its workload in steady-state condition is about 33.3% of the workload of the proportional replication algorithm. Meanwhile, more of one in a thousand peers has the desired requested rates in steady-state. Besides, the proposed algorithm has more stationary transient process.

Key word：
P2P network

deficit bandwidth

popularity

replication algorithm

replacement algorithm

View(8746)

Cell selection in energy harvesting enabled HetNets toward green 5G

DOI：10.16157/j.issn.0258-7998.180367

Author：Zhang Demin，Fu Yongli，Yang Kang

Author Affilications：Chongqing Key Lab of Mobile Communications，Chongqing University of Posts and Telecommunications，Chongqing 400065，China

Abstract：As the number of users in cellular mobile networks increases, cellular network changed from the previous single architecture to heterogeneous architecture, base station energy consumption also showed an exponential growth. Energy harvesting is an promising technology to save on-grid power and reduce carbon emission in wireless communications. The cell selection in the HetNets is the critical factor of affecting network performance. In this paper, from the energy harvesting, energy cooperation, heterogeneous networks and other aspects, a cell selection algorithm is proposed, which can compromise energy harvesting, energy consumption, load balancing and other criteria. The simulation results show that this algorithm can reduce the base station′s demand for grid energy, increase the possibility of user selection, and satisfy the fairness and validity. It′s applicable to 5G green communications.

Key word：
5G green communication

energy harvesting

energy cooperation

HetNets

cell selection

View(9264)

Resource allocation based on sleep strategy in heterogeneous networks

DOI：10.16157/j.issn.0258-7998.180616

Author：Zhu Jiajia，Wu Runze，Tang Liangrui

Author Affilications：School of Electrical and Electronic Engineering，North China Electric Power University，Beijing 102206，China

Abstract：This paper focuses on a multi-objective optimization approach for resource allocation based on sleep strategy, aiming to mitigate the interference and improve the energy efficiency in heterogeneous networks. Then it’s converted into a multi-objective optimization problem. The problem is solved by the nondominated sorting genetic algorithm version II. And the simulations show that the proposed approach is superior to the existing one in the aspect of mitigating the inter-tier interference between dense low power nodes and the macro base station and reducing energy consumption ratio of low power nodes.

Key word：
heterogeneous networks

sleep strategy

multi-objective optimization

resource allocation

nondominated sorting genetic algorithm version II

View(9594)

Computer Technology and Its Applications

Fully automated extraction of coronary main branches based on directional minimal path

DOI：10.16157/j.issn.0258-7998.181233

Author：Liu Liu，Xu Jin，Wu Pengfei

Author Affilications：Engineering Research Center of Wireless Communication，Nanjing University of Posts and Telecommunication，Nanjing 210003，China

Abstract：It has been a challenging but significant research topic to extract the centerlines of the coronary arteries of the CTA volume in clinical applications. A new method is proposed to full-automatically extract and recognize the centerlines of the major branches in this paper. This method, which is called the directional minimal path, originates from an improved minimal path algorithm. In this method, the cost function is redefined, and the starting point of the coronary is automatically detected. The proposed method is validated by extracting and recognizing the three major coronary branches, and the average overlap percentage was 81.0%.

Key word：
CTA image

coronary centerline

prior model

minimal path

coronary extraction

View(13154)

Smoke recognition based on the depth learning

DOI：10.16157/j.issn.0258-7998.180118

Author：Wang Tao，Gong Ningsheng，Jiang Guixiang

Author Affilications：College of Computer Science and Technology，Nanjing Tech University，Nanjing 211816，China

Abstract：Based on the second generation Artificial Intelligence Learning System(TensorFlow), this paper constructs a neural network to detect smoke images, and uses an improved motion detection algorithm to intercept the image of suspected smoke area. Combining with the PCA dimensionality reduction algorithm，the Inception Resnet V2 network model is trained to recognize the smoke characteristics under the TensorFlow platform. The algorithm realizes a large range of real-time fire detection alarm, and through experiments, it is proved that the whole detection process accurately identifies the smoke region in the video stream, which is more accurate and adaptive than the traditional smoke recognition method, and provides an effective scheme for the large range of fire smoke alarms.

Key word：
smoke detection

TensorFlow

PCA

convolutional neural network

motion detection

View(22430)

Design and implementation of monitor software applied on native platform

DOI：10.16157/j.issn.0258-7998.180540

Author：Zhang Shan，Liu Xiaokai，Wang Chao，Hao Kelin

Author Affilications：National Computer System Engineering Research Institute of China，Beijing 100083，China

Abstract：Currently, most monitor devices are developed based on foreign software and hardware. The core technology is controlled by others. So that this paper designs an monitor software which is applied on native software and hardware platform. By means of multi-database storage, the method of insert and query is optimized. A large number of parallel data is realized by using multi-threaded design. Using object-oriented program design method makes the data receiving layer, processing layer and display layer separated. Through testing by the actual environment, the carrying capacity, real-time and reliability of monitor is enhanced. At the same time, the software designed has good scalability and maintainability.

Key word：
native

monitor

multi-thread

data base

View(17306)

Target localization algorithm for strawberry robot-arm based on prior knowledge

DOI：10.16157/j.issn.0258-7998.174745

Author：Li Yanwen1，Cui Yang2，Li Lixin3

Author Affilications：1.School of Mechatronics Engineering，Harbin Institute of Technology，Harbin 150001，China； 2.School of Computer Science and Technology，Hangzhou DIANZI University，Hangzhou 310018，China； 3.Hik Group of China Electronics Technology Group Corporation，Hangzhou 310012，China

Abstract：This paper develops a strawberry robot-arm based on computer vision, and uses deep learning to localize targets for strawberry robot-arm. In order to speed up localization, the algorithm segments the image from strawberry robot-arm based on color prior knowledge to get strawberry proposals, then inputs proposals to convolutional neural network(CNN) pretrained and gets the probability for every proposal, and finally chooses the target with the highest probability as the localization result. Experiment results show that this algorithm can localize target in real time and with high precision.

Key word：
target recognition

deep learning

CNN

strawberry robot-arm

View(9302)

The navigation method and experimental research of air-ground robot cooperation

DOI：10.16157/j.issn.0258-7998.180065

Author：Liu Yi1，2，Che Jin1，2，Zhu Xiaobo1，2，Hang Li1，2，Hao Zhiyang1，2

Author Affilications：1.School of Physics and Electronic-Electrical Engineeing，Ningxia University，Yinchuan 750021，China； 2.Ningxia Key Laboratory of Intelligent Sensing for Desert Information(Ningxia University)，Yinchuan 750021，China

Abstract：Ground mobile robots and airborne robots have the obvious superiority of complementing each other, which plays an important role in improving the completion of missions. Therefore, more researchers have paid attention to those research in recent years. A typical model of the collaboration of air-ground robot is to provide the ground robot with path guidance information through the wide-range environmental awareness of the air robot. So firstly, a ground-based robot collaboration platform is established and the ground environment is perceived and identified by using airborne visual sensing information. Then, the path planning algorithm is used to plan the path and streamline the instructions. Finally, the ground robot performs machine vision tasks. The success rate of the experimental verification results is more than 80%, which verifies the feasibility of the method.

Key word：
air-ground robots

threshold segmentation

route plan

embedded

collaboration platform

View(14306)

Shadow detection with multi-feature fusion and MMSE optimization

DOI：10.16157/j.issn.0258-7998.174639

Author：Zhang Han1，Yan Huaiping1，Zhang Zhan2

Author Affilications：1.School of Computer Science and Information Engineering，Anyang Institute of Technology，Anyang 455000，China； 2.School of Electrical Engineering and Automation，Henan Polytechnic University，Jiaozuo 454000，China

Abstract：In order to reduce the interference of shadow to detection results of moving targets, a shadow detection method is proposed, which can be made as the post-processing step of motion detection methods. On the basis of the target list detected by motion detection methods, this method extracts the brightness, contrast and structural features of the pixels between adjacent frames for the three color channels R, G and B, and fuses the three types of features to generate the similarity measure. And the objective function is designed according to the MMSE(minimum mean square error) criterion. The optimal pixel segmentation thresholds are obtained by the optimal solution, to detect and eliminate the shadow pixels in the motion detection targets. The shadow detection experiments are carried out on the shadow subset of Changedetection.net dataset. The experimental results show that this method has high detection rate and low detection time.

Key word：
shadow detection

motion detection

minimum mean square error

optimization

multi-feature fusion

View(9178)

Power Supply Technology and Its Application

Study of pressing energy harvesting technology based on PZT cantilever beam

DOI：10.16157/j.issn.0258-7998.174389

Author：Wang Zhihua1，2，Chen Dongyang1，2，Yao Tao3，Lv Dianli1，2，Zhang Huijuan1，2

Author Affilications：1.State Key Laboratory of Reliability and Intelligence of Electrical Equipment，Hebei University of Technology，Tianjin 300130，China； 2.Province-Ministry Joint Key Laboratory of Electromagnetic Field and Electrical Apparatus Reliability， Hebei University of Technology，Tianjin 300130，China； 3.School of Mechanical Engineering，Hebei University of Technology，Tianjin 300130，China

Abstract：Based on the positive piezoelectric effect of piezoelectric ceramic(PZT), an energy harvesting device with cantilever beam is designed for the single pressing energy harvesting, which can be used for wireless switches and so on. The mathematical analysis of the response of PZT cantilever beam is carried out on the condition of single pressing. Through comparison of the performances, a cantilever beam is made of the selected P-81 piezoelectric plates which are superimposed on the both sides of the copper alloy sheet. Finite element analysis shows that the double piezoelectric plates connected in parallel can export higher power than that connected in series. However, the natural frequency has no relation with their connecting forms. It indicates that the connecting forms of the double piezoelectric plates have no effect on its strain. The harvester is simulated and tested under the single pressing condition with 10 kΩ resistance load. The calculated and tested maximum instantaneous powers are 28.93 mW and 27.46 mW, respectively. The energy harvesting circuit is designed according to the high impedance characteristic of the PZT at low frequency. A threshold switch circuit is used to control the discharge voltage. The experimental results show that the output electric energy of the energy harvesting circuit can reach 5.05 mJ. The energy can meet the requirements of driving LED, RF transmitter and so on.

Key word：
PZT

cantilever beam

energy harvesting

finite element analysis

circuit design

View(11830)

Study of a new type underwater bidirectional wireless charging system

DOI：10.16157/j.issn.0258-7998.180190

Author：Gao Xuefei1，Zhang Jian1，2，Li Jinlong1

Author Affilications：1．Key Laboratory of Power Electronics and Electrical Drives，Institute of Electrical Engineering， Chinese Academy of Sciences，Beijing 100190，China； 2.Beijing Engineering Laboratory of Electrical Drive System & Power Electronic Device Packaging Technology， Institute of Electrical Engineering，Chinese Academy of Sciences，Beijing 100190，China

Abstract：This paper presents a new type of power supply for information network. This scheme has adopted a new topology which can transmit electric energy bidirectional underwater. This function can realize both energy storage and external charging. In this paper, the compensating circuit has been confirmed based on the practical application condition, then the influence of the working frequency to the system has been analyzed, the influence of the eddy current loss in sea water environment has been analyzed and vivificated. At last the experiment of the wireless charging undersea has been completed. This paper has realized 83% of the transfer efficiency when the transmission distance is 15 cm. The result can achieve the technical indicator and verify the feasibility and effectiveness of the proposed strategy.

Key word：
bidirectional wireless charging

undersea environment technology

eddy current loss

View(9532)

Research on load regulation of Buck constant current converter

DOI：10.16157/j.issn.0258-7998.175180

Author：Li Yiming1，Wei Jincheng1，Guo Xiaoying2，Zhang Yufeng1，He Jinyu1，Li Yong3，Cao Taiqiang1

Author Affilications：1.School of Electric Information，Xihua University，Chengdu 610039，China； 2.School of Information and Electric Enaineerina，Panzhihua University，Panzhihua 617000，China； 3.Zhengzhou Power Supply Company，State Grid Henan Electric Power Company，Zhengzhou 450000，China

Abstract：There is a problem of poor load regulation in traditional LED power supply. In this paper, a control strategy of inductor current compensation based on the output voltage feedback is used to compensate the output current deviation caused by the load adjustment with the load voltage signal. In this paper, the factors affecting the load regulation of Buck constant current topological are analyzed theoretically. Combined with analysis of engineering application data, the control strategy of load regulation is optimized. Finally, the control method is verified on a LED power supply of 200 watts. The load regulation rate was reduced from 0.03%/V to 0.01%/V, which verified the correctness of the control methodproposed and achieved the effect of optimizing the load regulation rate.

Key word：
LED power supply

load regulation

inductance peak current

View(14194)

Study and design of an adjustable rail-to-rail output single-stage Buck converter

DOI：10.16157/j.issn.0258-7998.174140

Author：Tang Jianqiang，Zhou Fengxing，Mei Mingyang

Author Affilications：School of Information Science and Technology，Wuhan University of Science and Technology，Wuhan 430081，China

Abstract：As an important component of the cable fault detector, the output safety, stability and adjustable output range of the high voltage power supply play an important role in promoting the rapid detection of the cable fault point. In this paper, a wide range of adjustable single-stage Buck converter with rail-to-rail output characteristics is designed as a pre-regulator for the wide range of adjustable output high power supply. The closed-loop feedback compensation controller and software-configurable soft-start control method are analyzed and designed by using the unified circuit model of Buck converter. At the same time, the hardware circuit design of Buck converter based on TL494 is completed. Finally, simulation analysis and experimental test results show that this design of the Buck converter which can achieve the output voltage range from 10 to 300 V continuous adjustability has strong stability and high dynamic response. It can work continuously in 2 hours without exception.

Key word：
rail-to-rail adjustable output

buck converter

feedback compensation

TL494

soft start

View(11242)

Analysis of heat dissipation for power module based on spatial interpolation algorithm

DOI：10.16157/j.issn.0258-7998.180047

Author：Lu Di1，Xue Jingwei1，Liang Jianing2

Author Affilications：1.School of Mechanical and Electrical Engineering，Xi′an University of Architecture and Technology，Xi′an 710055，China； 2.Shenzhen Institute of Advanced Technology，Chinese Academy of Sciences，Shenzhen 518055，China

Abstract：In view of the difficulty in calculating the power module loss of the electric air conditioner and lack of heat dissipation design basis, the mechanism of heat production and heat dissipation is analyzed, the calculation method of actual loss value is proposed based on the spatial interpolation algorithm，mathematical model of heat transfer from a power module to a refrigerant is established, and two methods to improve the cooling efficiency is proposed. The thermal simulation analysis of the temperature distribution of power module under different thermal conductivity and base thickness is carried out by the finite element software. The theoretical analysis and simulation results are verified by experiments. The results show that the two proposed ways to improve the cooling efficiency can improve the thermal performance of power modules.

Key word：
electric air conditioner

power module

interpolation algorithm

heat dissipation

View(8050)

Power system harmonic detection based on cubic spline function and Rife-vincent self-convolution window interpolation FFT algorithm

DOI：10.16157/j.issn.0258-7998.182545

Author：Zhang Yingying

Author Affilications：Xinyang City，Henan Province，Quality and Technical Supervision，Inspection and Testing Center，Xinyang 464000，China

Abstract：When using the Fast Fourier Transform(FFT) to perform harmonic analysis on power systems, due to asynchronous sampling and data truncation, fence effect and spectrum leakage will occur, and accurate harmonic parameters cannot be obtained. In order to improve the accuracy of harmonic detection of asynchronous sampling, this paper proposes a Rife-vincent self-convolution window interpolation FFT algorithm based on cubic spline function. The Rife-vincent self-convolution window has low peak value and fast sidelobe attenuation, which can effectively suppress spectral leakage. The cubic spline function is used to approximate the amplitude ratio function, which can effectively suppress the fence effect and avoid solving high-order equations. The calculation accuracy is high. Through MATLAB simulation analysis, it is verified that the Rife-vincent self-convolution window interpolation FFT algorithm based on cubic spline function can effectively suppress spectrum leakage and fence effect and improve harmonic detection accuracy.

2018 No. 10

Publish Date：2018-10-06

ISSN：0258-7998

CN：11-2305/TN

Special Column-Artificial Intelligence

Key word： TensorFlow deep learning Convolutional Neural Network(CNN) digital recognition View(17044)

Key word： behavior recognition batch normalization long short-term memory neural network deep learning View(19166)

Key word： CapsNets finger vein recognition deep learning CNN View(11958)

Key word： deep learning CNN ZYNQ hardware-software-collaborative View(11762)

Key word： colorization convolutional neural network deep learning Squeeze-and-Excitation Networks(SENet) model Power Linear Unit(PoLU) View(10954)

Review and Comment

Key word： wireless sensor networks energy harvesting cognitive radio resource management View(9554)

Microelectronic Technology

Key word： CMOS-MEMS process high aspect ratio deep trench etching RIE combined with Bosch View(13814)

Key word： FPGA base-band shaping filter upper sampling arbitrary multiples ISI View(16504)

Key word： wavelet transform time-domain frequency-domain switched current View(9184)

Key word： MEMS inertial navigation system forgetting factors adaptive attitude calculation View(11910)

Key word： distributed power amplifier GaAs pHEMT View(11266)

Key word： logic obfuscation dummy contact domino logic hardware security View(10362)

Key word： truly random number generator thermal noise oscillator random tremble power consumption View(14542)

Embedded Technology

Key word： VxWorks 6.5 TRDP protocol stack CCU View(16554)

Key word： FPGA optical fiber vibration cross-correlation positioning View(18726)

Key word： real-time industrial ethernet EtherCAT pulse generator View(12248)

Key word： STM32 microcontroller OV2640 image sensor mean gray value erosion algorithm automatic detection View(7932)

Key word： pavement damage automatic detection extended Kalman filtering machine vision distance compensation View(8818)

Measurement Control Technology and Instruments

Key word： FPGA weak signal detection portable phase lock zoom View(28064)

Key word： ESD current waveform Adams prediction-correction algorithm wavelet denoising adaptive Kalman filter View(9902)

Key word： motion planning of robot artificial potential field angle of rotating speed vectors TAS-RRT View(16068)

Key word： personal navigation MEMS extended Kalman Filter zero-velocity detection and correction View(10344)

Key word： indoor positioning ZigBee wireless network fuzzy c-means weighted k-nearest neighbor View(14306)

Key word： RFID WSN mobile positioning distributed WSN fingerprinting View(11126)

Key word： virtual reality technology training immersive interactive View(13678)