15 July 2017, Volume 43 Issue 7
-
Select all|
-
Computer Engineering. 2017, 43(7): 1-8. https://doi.org/10.3969/j.issn.1000-3428.2017.07.001Existing methods for TV drama popularity prediction in video websites solely consider the single factor and most of them cannot predict before the premiere.This will lead to the video website making unreasonable decisions on the purchase of copyright and advertising with a few days’ time lag.To solve this problem,this paper proposes a method which can predict the TV drama popularity in video websites before its premiere.This method first uses the search data of TV drama such as name and actor comprehensively.Then the method calculates the earliest prediction time through time series analysis.Finally,based on multiple linear regression model,it gets the optimal feature and predicts the popularity.Experiments result shows that this method can use the thirteenth to eighteenth days’ Baidu search index of name and actor before the premiere to predict 30 days’ on demand quantity ranking after premiere for TV dramas launched on PPTV and YOUKU in 2014 and 2015.The Spearman correlation coefficient between the prediction rank and real rank reaches 0.943 7 on PPTV and 0.967 6 on YOUKU.The result shows that this method has a good prediction effect.
-
Computer Engineering. 2017, 43(7): 9-14. https://doi.org/10.3969/j.issn.1000-3428.2017.07.002Single Instruction Multiple Data(SIMD) aims at exploiting the data-level parallelism of multimedia and scientific calculation.The true dependence caused by reduction operation hinders exploring data-level parallelism.But different architecture and instruction set make reduction vectorization method based on vector machine unsuitable to SIMD.This paper presents a new approach of generating SIMD code with reduction.It presents a method to identify reduction operations,and conducts code generation using vector shift instructions.Experimental result based on SPEC and some other standard test sets shows the validity of this method,with 34% improvement in the vectorization acceleration ratio.
-
Computer Engineering. 2017, 43(7): 15-21. https://doi.org/10.3969/j.issn.1000-3428.2017.07.003
The existing researchon sensing information quality mainly focuses on the mobile nodes recruiting,selecting and sensing task allocating,lacking optimizing the execution of sensing tasks.This paper proposes Sensing Task Migrating(STM) method based on utility among the distributed heterogeneous mobile sensing devices which can collect sensing data collaboratively.The process of task execution is optimized to solve the contradiction between the heterogeneity of mobile devices and the qualty reauirement of sensing information.Experimental results show that,compared with random selecting algorithm and participant selecting algorithm based on multi-tasking,the algorithm improves the ratio of sensing data coverage and the ratio of sensing task finished.
-
Computer Engineering. 2017, 43(7): 22-28,37. https://doi.org/10.3969/j.issn.1000-3428.2017.07.004There exists some temporal databases and temporal analysis tools based on cluster-based computing systems.However,most of them are disk-oriented and performance degenerate rapidly when processing big data.This paper proposes a system which is based on Spark,and provides accessible and scalable temporal query scheme with large temporal data for users.Specifically,it extends Spark SQL parser to support SQL-like temporal operations.Besides,it uses the index manager based on Spark SQL which is proposed by SIMBA,and embeds optimization strategies in two aspects:global filtering and local temporal index.Depending on these optimization rules,the system achieves high throughput and low latency in temporal operations.Evaluation experiment results on temporal query efficiency and effectiveness show this system has improved temporal query performance over original Spark SQL in different factors.
-
Research on Minimum Cost Data Replica Distribution Based on Dynamic Planning in Cloud Storage SystemComputer Engineering. 2017, 43(7): 29-37. https://doi.org/10.3969/j.issn.1000-3428.2017.07.005
Traditional replicas distribution strategies consider less data transmission cost and constrainsd network structure lack of universality,a minimum cost of replicas distribution strategy using dynamic planning technology is put up to minimize the overhead of data management in the cloud storage system.It presents a comprehensive data management model including data storage,transmission and the update cost,compares their values and chooses the one with lower cost for replica placement.Experimental results show that the strategy can realize reasonable distribution of replica from global minimum cost view,which can reduce the overall data management cost,and effectively reduce the network transmission and the average response time,and therefore promote the development of cloud storage system.
-
Computer Engineering. 2017, 43(7): 38-43. https://doi.org/10.3969/j.issn.1000-3428.2017.07.006Cache coherence protocol impacts the system performance and the demand of bandwidth.Snoopy protocols and directory protocols are widely used in modern server systems,but the former needs large bandwidth while the latter has long latency,so they are not suitable for domestic server CPU used in the scenario that the bandwidth is relatively small and latency is relatively long.To solve this problem,this paper proposes a hybrid coherence protocol based on Token protocol and directory protocol,and uses Simics and GEMS to construct a system in which multi-channel Chip Multi-processors(CMP) are directly connected.It then runs SPLASH-2 benchmark to test related performance.The experimental results show that the hybrid protocol has better performance than the directory protocol.Compared with the Token protocol,the hybrid protocol significantly reduces the demand of inter-chip bandwidth at relatively little cost of performance,and has better performance in the system that the bandwidth is not sufficient.
-
Computer Engineering. 2017, 43(7): 44-47. https://doi.org/10.3969/j.issn.1000-3428.2017.07.007A lot of debug work is required when implementing specific functions on Software Communication Architecture(SCA) waveform components,and the usual debug step is to extract the output data of each process for analysis and modify the error code according to the analysis result.However,the debug method is inefficient and the data extraction process is complicated.To this end,this paper designs the AnalysisCom waveform components for data analysis based on Matlab engine technology and curses graphics library.The component is designed and developed in accordance with the SCA specification.It can avoid the duplication of the extracted data during debug and directly call Matlab to analyze data in the process of the waveform running.Test results show that,AnalysisCom components can locate where the error occurred fast and accurately while the waveform is running,complete the verification of the component’s functionality,and achieve the goal of shortening the development cycle.
-
Computer Engineering. 2017, 43(7): 48-53. https://doi.org/10.3969/j.issn.1000-3428.2017.07.008Since there is no requirements analysis tool for on-board train control software,requirements analysis researchers are unable to get expected data.Aiming at this problem,this paper proposes a requirements analysis method based on model extraction and designs a Control Software Requirements Analysis(CSRA) tool.It constructs an executable model by extracting the requirement items of requirements description documents,uses data flow method to analyze the model and obtain the Variable Dependency Diagram (VDD) and State Transition Diagram (STD).By using dynamic analysis,it gets the expected output data based on the requirements description documents and compares the data with real output data to generate an visual analysis report of the control software.Experimental results show that the proposed method can help requirements analysis researchers quickly find errors in software implementation,which provides a guarantee for the safe running of trains.
-
Computer Engineering. 2017, 43(7): 54-59. https://doi.org/10.3969/j.issn.1000-3428.2017.07.009
For the matching of multiple locks and multiple keys,a low-power intelligent key which is applied to the active authentication with embedded radio frequency identification is designed.The management of multiple locks is realized by means of authorized registration and active two-way authentication,where password dictionary encryption and quasi dynamic password index are used to ensure that the authentication process is reliable.According to the low power requirement of the key,a 1 s periodic sleep with 0.005 s wakeup mechanism is proposed to reduce the power consumption of authentication card as much as possible.Experimental results show that,the power consumption of the intelligent key sleep mode is less than 2 μA,and it meets the requirement of matching and security authentication between multiple keys and multiple locks.
-
Computer Engineering. 2017, 43(7): 60-63,69. https://doi.org/10.3969/j.issn.1000-3428.2017.07.010Optimizing system energy cost is important for extending the life cycle of embedded system.Although existing scheduling algorithms can optimize the task scheduling to a certain extent,but there are still some problems in the system energy cost and task scheduling stability.In order to solve these problems,this paper puts forward a new task scheduling algorithm.It optimizes the system costs and increases the task scheduling stability through improving the system resource utilization rate and reducing the number of bumps in task scheduling.Experimental results show that,compared with LSF,EDF and HVF algorithms,in the same task scheduling environment,the proposed algorithm can achieve better balance between system resource utilization rate and task scheduling stability.
-
Computer Engineering. 2017, 43(7): 64-69. https://doi.org/10.3969/j.issn.1000-3428.2017.07.011Aiming at the problems that the guidance mode of traditional Automated Guided Vehicle(AGV)has weak anti-interference,low flexibility,and high cost,this paper proposes a new cloud-based guidance mode,and designs and implements the core Cloud-based Guided Platform(CBGP) providing the guidence.The CBGP server provides access to the platform,communication and management functions,MapReduce computing framework and HDFS file system provide distributed computing and storage.Experimental results show that the CBGP platform can apply a good path planning algorithm to provide the appropriate path and complete the task of cloud guided AGV.
-
Computer Engineering. 2017, 43(7): 70-74. https://doi.org/10.3969/j.issn.1000-3428.2017.07.012Aiming at the performance loss of BESIII Offline Software System(BOSS) on Kernel Virtual Machine(KVM) platform,this paper gives related optimization suggestions combining BOSS operation characteristics,and studies the factors which lead to performance loss through quantitative analysis and test results,and determines performance loss.It explores the performance loss of multiple clients.BOSS job performance test on client machina and physics machina,results show that the performance loss optimize simulation job is decreased to 1.1%~1.6%,the performance of reconstruction job is increased by 2.6%~4.5%,the performance of analysis job is increased by 7%~18.7%.
-
Computer Engineering. 2017, 43(7): 75-79. https://doi.org/10.3969/j.issn.1000-3428.2017.07.013To further improve the networking capability of ZigBee Ad Hoc network,this paper researches ZigBee Ad Hoc network and its route algorithms.Firstly,a network is created by the ZigBee technology.Then,the energy balance of network nodes and the convergence rate are taken into account when the network updates its routing path.Finally,an improved Genetic Algorithm(GA) is applied to search the global better solution,and a Particle Swarm Optimization(PSO) algorithm is adapted to quickly search the global optimal solution.Simulation results in NS2 show the superiority of GA-PSO algorithm in network lifetime and propagation delay by comparing with traditional AODVjr route algorithm and the route algorithm based on GA.It is more suitable for the larger complex network.
-
Computer Engineering. 2017, 43(7): 80-85. https://doi.org/10.3969/j.issn.1000-3428.2017.07.014At present,the wide range of sparse Wireless Sensor Network(WSN) uses VHF communication,but the single channel resources cannot satisfy the requirement of communication among nodes in the network.For this purpose,this paper puts forward an adaptive cognitive frequency wireless sensor node device.It designs processor module,wireless communication module,solar energy power supply module,and sensor module,and defines the dynamic channel allocation protocol software,to realize the adaptive frequency communication of channel cognitive function.The experimental results show that the device can undertake continuous accurate wireless communication over a long distance,and complete the adaptive frequency communication.
-
Computer Engineering. 2017, 43(7): 86-89. https://doi.org/10.3969/j.issn.1000-3428.2017.07.015In view of the hybrid wireless communication system of eHRPD in 3G and LTE-FDD in 4G,the cell of each network type will overload with the increase of users.Aiming at this problem,this paper proposes a new load balancing algorithm.It refers to the network resource utilization rate to vertically switch users between different systems,so as to balance the load of the system in the hybrid network.Simulation results show that,for the LTE-FDD and eHRPD overlap coverage system,the proposed algorithm can enhance the average throughput while the wireless link lose rate is unaffected.Moreover,the best handover thresholds can be obtained which make the throughput largest.
-
Automatic Measurement and Analysis of Antenna Radiation Characteristics for Wireless Sensor NetworksComputer Engineering. 2017, 43(7): 90-99. https://doi.org/10.3969/j.issn.1000-3428.2017.07.016Antenna radiation characteristics of the antenna are important parameters in Wireless Sensor Networks(WSNs),but they are often overlooked or simplified because the antenna radiation characteristics measurement process is professional,expensive and time-consuming,and the characteristics are affected by many parameters.Aiming at this problem,this paper presents an automatic antenna measurement system and data processing program for WSNs,which can quickly and easily implement measurement and data processing of the antenna radiation characteristics.Furthermore,radiation characteristics of fourteen typical antennas in WSNs on different planes and the influence of metal planes on antenna directivity are measured and analyzed.At last,the characteristics and target application scences of all antennas are discussed.This study is expected to provide suggestions and references for antenna selection and installation in WSNs for different applications.
-
Computer Engineering. 2017, 43(7): 100-109. https://doi.org/10.3969/j.issn.1000-3428.2017.07.017In Passive Sensing Network(PSN),Network Coding (NC) based IPv6 packets delivery schemes mostly use fixed number of fragments,which leads to long transmission delay.In order to reduce the transmission delay,an IPv6 Packet Delivery Scheme called IPDS-DF which can be applied to PSN is proposed considering the characteristics of PSN nodes.The Successful Transmission Probability(STP) is derived based on the BER model of Offset Quadrature Phase Shift Keying(OQPSK).Furthermore,the average transmission delay,the average energy consumption and the end-to-end Packet Delivery Ratio(PDR) are also derived.For minimizing the transmission delay,the number of fragments in each hop of the route are dynamically optimized.Experimental results indicate that the IPDS-DF outperforms the existing Reliable IPv6 Packet Delivery Scheme(RIPDS) in terms of average transmission delay and average energy consumption.
-
Computer Engineering. 2017, 43(7): 110-118,123. https://doi.org/10.3969/j.issn.1000-3428.2017.07.018
Current erasure code repair methods have problems such as high expenses and low efficiency.Aiming at these problems,this paper brings forward a multi-node failure repair method with low cost.By means of network distance based node choosing method,available bandwidth between nodes is enhanced.Data transmission method with multithreading and pipeline is adopted for the purpose of improving multi-node failure repair efficiency.Central node based multi-node repair method is utilized for decreasing multi-node repair cost.Experimental results show that,compared with current Star-like Structure Based Serial Repair(SSR) strategy,Tree-like Structure Based Serial Repair(TSR) strategy and Minimum Storage Regenerating(MSR) codes,the proposed method has higher multi-node repair efficiency,and it can reduce the average repair time by 25%,16% and 20%,respectively.
-
Computer Engineering. 2017, 43(7): 119-123. https://doi.org/10.3969/j.issn.1000-3428.2017.07.019To improve the outage performance of cognitive radio relay network,this paper proposes a power allocation algorithm minimizing outage probability by analyzing the system outage probability and user power allocation conditions.It presents the mathematical optimization equations minimizing outage probability of under master user interference constraint and total power constraint.Then,according to the relationship between the user’s maximum transmit power and interference level threshold,it proposes three optimal power allocation schemes respectively to deal with three types of cases,including the distribution of power not exceeding the interference level threshold,the relay node power limited by the interference level threshold,and all nodes limited by the interference level threshold.Finally,the optimal value is solved based on Karush-Kuhn-Tucker(KKT) conditions.Experimental results show that compared with similar power allocation methods,the proposed algorithm has obvious performance gain promotion.And the effectiveness of the adopted power allocation schemes is proved the most effective in enhancing the system outage performance when the total power or the interference level is changing.
-
Computer Engineering. 2017, 43(7): 124-128,135. https://doi.org/10.3969/j.issn.1000-3428.2017.07.020In massive Multiple-Input Multiple-Output(MIMO) system,In-Band(IB)Full-Duplex(FD) wireless backhaul scheme can reuse the radio access spectrum and greatly improve the network performance.With this consideration,this paper proposes a massive MIMO heterogeneous cellular network model based on IB FD wireless backhaul. By assuming that the network elements are subject to independent Homogeneous Possion Point Process(HPPP)and using the stochastic geometry theory,it researches the Signal-to-Interference-Noise Ratio(SINR) coverage probabilities of the macro cell downlink and small cell uplink transmission respectively,and analyzes the impacts of different parameters on the coverage performance.Simulation results show that the IB FD wireless backhaul scheme can effectively improve the spectral efficiency of the system,but it cannot improve the SINR coverage probability of the macro cell and small cell networks.
-
Computer Engineering. 2017, 43(7): 129-135. https://doi.org/10.3969/j.issn.1000-3428.2017.07.021As existing approaches for verifying Multipath Transmission Control Protocol(MPTCP) performance are based on simulation environment,which cannot effectively verify the performance of MPTCP in realistic network,an experimental platform is constructed in this paper based on real network environment according to the constraint principles of MPTCP.Moreover,it implements OLIA,Balia,MPVeno and wVegas and conducts a variety of experiments scenarios based on the proposed platform to evaluate the performance of the above-mentioned four congestion control algorithms in throughput,load balance and fairness.Experimental results show that,MPVeno performs best in throughout.And with different subflow link congestion,MPVeno can acheive more effective usage of network source.As for the fairness index,Balia and wVegas perform better than MPVeno and OLIA.
-
Computer Engineering. 2017, 43(7): 136-140,145. https://doi.org/10.3969/j.issn.1000-3428.2017.07.022Generally,with higher average energy per bit to noise power spectral density ratio at the receiver,the packet can be more successfully detected.However,this situation quires more energy,lowering Energy Efficiency(EE).Based on the configurability of Cognitive Radio(CR),this paper studies the trade-off between Spectral Efficiency(SE) and EE from the angle of contacting with practical application more closely.It proposes a general metric Spectral/Energy Efficiency(SEE ) scheme to facilitate the analysis which quantifies the preference of SE or EE.Closed-form solutions for symbol transmission energy and the length of information bits per frame are obtained for various combined modulation and channel coding modes.The closed-form solutions further facilitate the adaptability of cognitive radio considering both SE and EE in various scenarios.Simulation results show that the scheme is capable to perform balanced trade-off between SE and EE,it can obtain better performance than simply optimizing any of the two metrics.
-
Computer Engineering. 2017, 43(7): 141-145. https://doi.org/10.3969/j.issn.1000-3428.2017.07.023The problem of pilot contamination is the main factor which influences the performance of Massive Multiple Input Multiple Output(MIMO) system.In order to solve this problem,this paper proposes a new pilot contamination mitigation algorithm in massive MIMO.Firstly,a two-dimensional compression sensing model in spatial and temporal domains is built.Then,pilot signal reconstruction is carried out by using Orthogonal Matching Pursuit(OMP) algorithm.The simulation results show that compared with precoding algorithm based on Minimum Mean Square Error(MMSE),under the condition of the same system demand,this algorithm can further reduce the number of pilots to mitigate pilot contamination effectively and improve the whole performance of massive MIMO system.
-
Computer Engineering. 2017, 43(7): 146-150,155. https://doi.org/10.3969/j.issn.1000-3428.2017.07.024Initial weight and threshold are random in the process of training Back Propagation Neural Network(BPNN).Moreover,BP algorithm is easy to fall into local optimum,so the training of BP neural network is uncertain to get the optimal network structure.To solve this problem,this paper proposes a spectrum prediction method of HS-BP neural network.Using the global optimization capability of harmony search algorithm,it can get the optimal initial weights and thresholds of the BP neural network,ensuring optimal BP neural network structure of training,ultimately,it can use the optimal BP neural network structure of training to predict the spectrum.Simulation results show that the algorithm can improve the accuracy of spectrum prediction and further improve the utilization of the spectrum.
-
Computer Engineering. 2017, 43(7): 151-155. https://doi.org/10.3969/j.issn.1000-3428.2017.07.025In relayed transmission systems,power allocation algorithm based on Decode and Forward(DF) increases error in case of incorrect decoding of the information symbols,and the one based on Amplify Forward(AF) amplifies the noise.In order to solve the problem,a Cognitive Radio(CR) Orthogonal Frequency Division Multiplexing(OFDM) power allocation algorithm based on Adaptive Relay(AR) is proposed in this paper.The proposed algorithm adaptively selects proper relayed forward scheme according to the Signal Noise Ratio(SNR) to allocate power.The objective is to maximize the system capacity while the interference introduced to the Primary Users(PU) is below the tolerated limit.Simulation results show the proposed AR algorithm achieve higher capacity compared with DF and AF,whenever in low SNR areas or in high SNR areas.
-
Computer Engineering. 2017, 43(7): 156-162. https://doi.org/10.3969/j.issn.1000-3428.2017.07.026Jamming attacks may cause the change of node state transition rule.Based on the statistical analysis of node state transition,this paper puts forward an improved jamming detection method.At the learning phase,the jamming detection judgment thresholds and jamming type judgment thresholds of the proportion of node state time are extracted through learning from samples collected from jamming and jamming-free scenarios.At the detection phase,these corresponding decision thresholds are compared with the proportion of node state time to detect the jamming attacks and determine the jamming types.The method of weighted detection confidence is proposed to further improve the detection rate and reduce the false alarm rate.Simulation results on NS3 validate that the proposed algorithm can accurately detect the typical on-demand and continuous jamming attacks with low false alarm rate.
-
Computer Engineering. 2017, 43(7): 163-169. https://doi.org/10.3969/j.issn.1000-3428.2017.07.027The existing secret sharing scheme in the vector space cannot achieve the secret share update or the dynamic addition and deletion of members,and it is hard to ensure the existence of trusted center,so a multi-secret sharing scheme without trusted center is proposed.The secret share is negotiated by all participants without trusted center,and anyone can verify the correctness of distributed secret shares and updated ones with the properties of bilinear pairings.A new public function is constructed to express the shared secret,and participants can reconstruct multiple secrets by the public function.Addition and deletion of members are determined by other members jointly,and the way that old members assist new members for generating their secret shares is adopted,which achieves dynamic management of members.The correctness and security of the scheme are proved.Compared with other vector space secret sharing schemes,this scheme has higher security,efficiency and dynamic property.
-
Computer Engineering. 2017, 43(7): 170-174,181. https://doi.org/10.3969/j.issn.1000-3428.2017.07.028Aiming at the imformation security assessment problem of security protection system of Internet Finance,based on the specific requirements of 《Internet Financial Networks and Information Security Technical Guidance》and the elements of P2DR2 security model,this paper puts forward an index system of information security evaluation for Internet Finance.Analytic Hierarchy Process(AHP) is adopted to determine index structure and weight.Combined with fuzzy comprehensive evaluation method,the overall evaluation results of Internet financial information are given.The example application results show that the proposed index system can effectively evaluate the information security level of Internet Finance,and it has stronger practicability.
-
Computer Engineering. 2017, 43(7): 175-181. https://doi.org/10.3969/j.issn.1000-3428.2017.07.029In order to distinguish detection attack categories pertinently and improve the classification accuracy of Intrusion Detection System (IDS),a hierarchical attributes reduction model used for IDS is proposed.This model reduces attributes by adopting dual structure of culture algorithms and combined with rough sets as well as genetic algorithm.Firstly,the data is preprocessed and divide it into hierarchies,which forms the rule of decision subset fD.Secondly,using cultural algorithm,knowledge is updated in belief space,and evolving data of hierarchical evaluation knowledge is introduced into population space.Thirdly,the optimal subser of each layer is acquired by using rough sets and genetic algorithm which can evolve knowledge and reduce attributes in belief space.Finally,model performance is verified by designing Bayes hierarchical classifier.Experimental results show that the algorithm can improve the accuracy of the Bayes classification before attributes reducting to 98.21%,and it is better to identify the intrusion categories whose traffic characteristics is not obvious such as R2L and U2R.
-
Computer Engineering. 2017, 43(7): 182-186,192. https://doi.org/10.3969/j.issn.1000-3428.2017.07.030Aiming at the problem of poor defensiveness of the current popular chaotic scrambling encryption algorithm, this paper proposes a new quantum chaos image encryption algorithm, which uses quantum Logistic chaotic sequence and bit scrambled image for encryption.Global scrambling of each pixel’s bits of plaintext is done, and encryption operation of the intermediate ciphertext is carried out to obtain the encrypted image.The algorithm solves the problems of bad randomness and few control parameters of Logistic chaotic system by introducing a quantum chaotic system.Experimental results show that this algorithm abandons XOR operation used in traditional encryption algorithms,and it performs better in the statistical characteristics, anti-attack performance and other aspects.
-
Computer Engineering. 2017, 43(7): 187-192. https://doi.org/10.3969/j.issn.1000-3428.2017.07.031To effectively solve the classification problem of imbalanced distribution of Chinese microblog sentiment data, this paper presents a classification method which combines affinity propagation algorithm, Word2vec technology and Conditional Random Field(CRF) model.It clusters the data of microblog by Affinity Propagation(AP) algorithm, and the majority class of samples are divided into several clusters according to the similarity measure for maximum distance between classes and minimum distance within class.The training set with balanced emotional tendency is constructed by using the undersampling technique.Word2vec is used to obtain the texts with the highest sentiment similarity to expand the microblog sentence and increase the sentiment information.The CRF model is used to calculate the label sequence of the balanced and extended training set.It is also possible to classify the emotional tendency of microblog accurately when the distribution of emotional tendency is unbalanced.Experimental results demonstrate that the proposed method achieves better results than ACRF method, CRF method and SCRF method in recall and G-mean evaluation criteria.
-
Computer Engineering. 2017, 43(7): 193-197,202. https://doi.org/10.3969/j.issn.1000-3428.2017.07.032Partial Least Squares(PLS) method adopts Principal Component Analysis(PCA),it cannot express the nonlinear characteristic,and the prediction accuracy is low in the nonlinear data.Based on this,an analysis and predicting method combining Restricted Boltzmann Machine(RBM) with PLS is proposed.The method can extract nonlinear structure of feature space by RBM and replace the components in PLS with the extracted components,forming a model which can adapt to nonlinear dose-effect relationship.Experimental results show that the method can well reflect the nonlinear characteristics of data.
-
Computer Engineering. 2017, 43(7): 198-202. https://doi.org/10.3969/j.issn.1000-3428.2017.07.033The performance improvement of the current Bat Algorithm(BA) lacks rigorous convergence theory, so that the blindness caused by the improvement of the algorithm does not have general theoretical significance.To solve this problem, this paper starts from the perspective of mathematical probability and the state transition of the BA meeting the Markov process.Through the establishment of a reasonable Markov chain model, the transfer behavior of individual status of bats is studied.It proves that the bat population state space is reducible and homogeneous, and the BA satisfies the convergence criterion of stochastic algorithm theoretically.It can converge to the global optimal solution with the probability of 100%.
-
Computer Engineering. 2017, 43(7): 203-210,216. https://doi.org/10.3969/j.issn.1000-3428.2017.07.034Most classifiers prefer to classify high-dimensional data based on the characteristics of low-dimensional subspace.However,its performance suffers from overlapping among different classes in the single subspace.To address this issue,a classification method using manifold learning based neural network is proposed.This method uses nonlinear embedding to attain the subspace of each class,and nonlinear embedding criterion to optimize the parameters of each basis function auto-associative neural network.Experimental results show that this method is more robust and has higher accuracy than traditional methods.
-
Computer Engineering. 2017, 43(7): 211-216. https://doi.org/10.3969/j.issn.1000-3428.2017.07.035Aiming at the deficiency of traditional Hidden Markov Model(HMM) in solving part-of-speech tagging,this paper proposes an improved second-order HMM.This model can better connect with contextual information,making part-of-speech tagging of the diagnosis of Traditional Chinese Medicine(TCM) accurate.Method of taking scale factor is proposed to solve the array of underflow and the new word processing method is given in the course of training.Experimental results show that,compared with traditional model,the improved second-order HMM has higher accuracy in part-of-speech tagging.
-
Computer Engineering. 2017, 43(7): 217-222. https://doi.org/10.3969/j.issn.1000-3428.2017.07.036Traditional keyword extraction is time and labor consuming.In order to automatically extract high quality keywords from the given document,an automatic keyword extraction method is proposed.TopK scored words are selected as keywords of a document after ranking each candidate phrase combining the topic information with statistical features of documents and phrases.Experimental results show that this method outperforms the existing methods in accuracy,recall rate and F score.It can effectively extract keywords from documents.
-
Computer Engineering. 2017, 43(7): 223-228. https://doi.org/10.3969/j.issn.1000-3428.2017.07.037Aiming at the characteristics of the hydrological monitoring video images,such as the high Signal to Noise Ratio(SNR),obvious color differences among targets and the specific location of target regions,an improved method for segmenting the hydrological image is proposed.By dividing the HSV color model into quantization intervals,the three-dimensional color information is changed into a one-dimensional array.To optimize the segmentation results based on the color only,the secondary quantizing of the Value component of HSV model takes place.Based on this,the current water level is obtained by using the improved region growing method.Finally,the segmentation of hydrological image is realized.The experimental results show that the method can segment the target area quickly and solve the connectivity and proximity problems between pixels in the hydrological image.
-
Computer Engineering. 2017, 43(7): 229-233. https://doi.org/10.3969/j.issn.1000-3428.2017.07.038In order to improve the processing speed and denoising effect of salt and pepper noise images,this paper presents a serpentine scanning filtering algorithm.First,in the distribution range of noise,it calculates a threshold to denoise by using the OTSU algorithm.And then,it searches a gray level whose difference with its neighbor is maximum in this range.Finally,it takes the average of the two gray levels and uses it as the segmentation point of signal and noise.Experimental results show that the algorithm can remove salt and pepper noise effectively,quickly,and accurately.Compared with median filtering algorithm,this algorithm preserves details very well,especially for images corrupted by high density noise.
-
Computer Engineering. 2017, 43(7): 234-238. https://doi.org/10.3969/j.issn.1000-3428.2017.07.039Aiming at the shortcomings such as time consuming,high complexity and sensitivity to noise,illumination and the sudden moving of objects in the traditional mixed Gaussian model algorithm for object detection,an improved moving object detection algorithm is presented in this paper.Based on the mixed Gaussian model and six-frame difference algorithm,the fundamental outline of detecting target is established.The accuracy of the algorithm is improved by using an adaptive selection strategy with the update rate for different regions.By means of the morphological operations,the residual noises are removed and the integrity of the detected target is improved through connectivity test.Then the binarization detection results of moving target with clearer and more complete outline are obtained.Simulation results show that the preposed algorithm not only improves the real-time performance of object detection,but also solves false detection problems caused by the change in object states,the environmental noise and the illumination changing,etc.
-
Computer Engineering. 2017, 43(7): 239-243,249. https://doi.org/10.3969/j.issn.1000-3428.2017.07.040
In order to improve the image classification accuracy,this paper proposes a Non-negative Elastic Net Sparse Coding(NENSC)algorithm.This algorithm combines the advantages of non-negative sparse coding and elastic net algorithm.It introduces an l2norm regularization term to the objective function of Sparse Coding(SC) optimization model and non-negative constraints to coding coefficients are applied.The proposed algorithm combined with Spatial Pyramid Matching(SPM) model is applied to image classification.Experimental results show that,compared with the traditional sparse coding algorithm,the proposed algorithm not only increases the prediction capability and effectiveness of the coding,but also makes the similar feature descriptors similar after coding and improves the stability of the coding,it has higher classification accuracy.
-
Computer Engineering. 2017, 43(7): 244-249. https://doi.org/10.3969/j.issn.1000-3428.2017.07.041In order to solve the problem of large amounts of computation in the global stereo matching algorithm,this paper proposes an improved algorithm by introducing global error energy function.It considers global error energy function as the cost of stereo matching,carries out region growing by leaps and bounds and gets the disparity map by jumping to obtain error energy function.The disparity map is smoothed by using average filter and the tolerance coefficient which will affect the size of filtering threshold is set,making it more suitable for human eyes.The tolerance coefficient is selected adaptively for different color and pixel stereo image pairs to get optimal filtered disparity map.Experimental results show that,on the basis of maintaining the accuracy,the improved algorithm can reduce computing time and improve real-time performance of matching.
-
Computer Engineering. 2017, 43(7): 250-252,260. https://doi.org/10.3969/j.issn.1000-3428.2017.07.042In order to reduce the high false alarm rate on foreground detection in dynamic scenes,this paper proposes an improved foreground detection algorithm,considering the high temporal redundancy of background images in the video sequence.Firstly,the Micro Video Block Element(MVBE) groups are formed by the sliding window technique and the redundancy values of each MVBE are computed by Spatio-Temporal Sum of Absolute Difference(STSAD) method.Secondly,The MVBE groups of low redundancy are modeled by Dynamic Texture(DT),and the DT parameters are estimated by Batch-Principal Component Analysis(Batch-PCA).Thirdly,the saliency value of every pixel in the MVBE is calculated by the observability measure method to realize foreground detection.Experimental results show,compared with the KDE,GMM algorithm,the algorithm has better inhibitory effect in dynamic scenes and a robust foreground detection result is obtained.
-
Computer Engineering. 2017, 43(7): 253-260. https://doi.org/10.3969/j.issn.1000-3428.2017.07.043In order to improve the detection accuracy and efficiency of infrared dim small target under complex background interference and low signal to noise ratio,a fast detection algorithm for infrared dim small targets based on local weighted fusion feature and classification 2D Ostu segmentation is proposed.Firstly,based on Top-Hat operator,the infrared background filtering mechanism is constructed by considering the gray characteristic difference between the background and the infrared dim small target,so that the infrared dim targets is highlighted from the background area.Then,the local weighted fusion feature is defined by the entropy value of the image to finish the rough positioning of dim small targets.The fractal dimension of all pixels in coarse positioning region is calculated by using fractal theory to complete the fine positioning of the dim small target.Finally,the feature classification 2D Ostu segmentation mechanism is defined to accurately detect the dim small target.The experimental results show that compared with the current infrared dim small target detection algorithm,the proposed algorithm has higher detection accuracy and shorter detection time.
-
Computer Engineering. 2017, 43(7): 261-267,273. https://doi.org/10.3969/j.issn.1000-3428.2017.07.044In order to solve the problem that it is difficult to detect the contour of the image based on the shape of the target,it can not take into account the global and local saliency,and the retrieval accuracy is not high.By introducing anisotropic filtering,design graphics saliency detection operator,to smooth the noise and retain high curvature features,its shape was accurate detection,considering the total shape of the contour points and the significant position,based on the shape of centroid,defined relative angular position calculation model,and combined with curvature function,determine the representation of each significant point values will shape the global features embedded into the local details,combined with dynamic programming algorithm to construct the shape significant similarity measurement model,feature point matching,complete graphics retrieval.The test results show that the proposed algorithm has higher retrieval precision and stronger robustness than the shape based image descriptor.
-
Computer Engineering. 2017, 43(7): 268-273. https://doi.org/10.3969/j.issn.1000-3428.2017.07.045In order to suppress Pseudo-Gibbs phenomena and remove false edges in image magnification of Wavelet domain,an image magnification algorithm in Wavelet domain is presented.After the original image is decomposed by one level Wavelet transform,an accurate location and placement method of high frequency coefficients in magnified high frequency subbands is given according to the orientation characteristic of high frequency subbands.An edge direction based interpolation method is proposed to determine the value of high frequency coefficient at the blank spaces of pixels.The magnified image is obtained by inversing Wavelet transform on the above magnified three high frequency subbands,and the magnified low frequency subband generated by linear transform of the original image.Experimental results show that the proposed algorithm can effectively remove the Pseudo-Gibbs phenomena due to the shift variance characteristic of wavelet transform,and avoid the generation of false edges.
-
Computer Engineering. 2017, 43(7): 274-280,287. https://doi.org/10.3969/j.issn.1000-3428.2017.07.046Aiming at the poor tracking performance problem of using single sensor in complex background,a fusion tracking algorithm for infrared-visible targets based on sparse representation and L1-APG is proposed.Firstly,the target models of infrared and visible light are built respectively by using sparse representation method,and the optimization problem is constructed by minimizing their joint reconstruction error.Secondly,the optimization problem is solved by employing L1-APG algorithm.Finally,the computational complexity of the algorithm is further reduced by using the minimum error boundary constraint,and dynamic updates of the infrared and visible target templates are achieved.Test results of infrared and visible image sequences demonstrate that the proposed algorithm performs better than the traditional single source tracking algorithm in dealing with target occlusion,target rotation,target intersection,and illumination change.
-
Computer Engineering. 2017, 43(7): 281-287. https://doi.org/10.3969/j.issn.1000-3428.2017.07.047To solve the problem of over-fitting and limited generalization ability during sensor fault diagnosis by traditional neural network,a fault diagnosis method for aerial sensor based on deep belief network observer is proposed.Shallow layer neural network is replaced by deep belief network.On the basis of optimizing network structure,the recurrence formula of selecting hidden layer nodes is proposed to build deep belief network state observer.Flight data is used to train deep belief network observer during offline training.Output of the observer is compared with actual output to judge the fault types and three methods of fault isolation and signal reconstruction are proposed during online diagnostics.Simulation results show that compared with BP neural network observer,the proposed method can diagnose and isolate faults and reconstruct signals with rapidity and high accuracy.
-
Computer Engineering. 2017, 43(7): 288-292,297. https://doi.org/10.3969/j.issn.1000-3428.2017.07.048For speech signals in low Signal-to-Noise Ratio(SNR) environment,residual background noise is large when using the traditional spectral subtraction method.Aiming at this problem,this paper puts forward an improved speech enhancement algorithm based on auditory masking effect.By combining the human ear auditory masking properties with power spectrum subtraction method,it puts forward a time domain recursive average algorithm to estimate noise.It makes spectrum subtraction for the speech signal with noise.From the perspective of hearing,it uses the estimated speech signal power spectrum to calculate the masking threshold.Finally,it introduces power correction coefficient spectral spectrum noise reduction coefficient to enhance speech signal with noise.Obtained by Matlab 2012b simulation,the experimental results show that the proposed algorithm can better suppress the background noise and improve the speech quality under low SNR conditions,and the SNR of the output signal is improved by about 5% compared with the improved adaptive filtering algorithm.
-
Computer Engineering. 2017, 43(7): 293-297. https://doi.org/10.3969/j.issn.1000-3428.2017.07.049For automatic driving and advanced driving assistance task based on computer vision,the real-time detection of travelable area in front of the vehicle is necessary.The travelable area detection is an image segmentation problem,and the mainstream Scene Parsing scheme is based on deep learning model.In general,the deep learning framework is not real-time and cannot be applied in automatic driving task or advanced driving assistance task until now.Aiming at this problem,in this paper,an Appearance and 3D feature fusion detection framework is proposed,which implements two-stage process from coarse-grained segmentation to fine-grained segmentation.The coarse-grained segmentation is the fast estimation based on RANSAC fast plane,and the fine-grained segmentation is the conditional random field model based on the road estimation probability from coarse-grained segmentation,which uses super pixels to accelerate.Experimental result in the Cityscape data sets shows that the proposed framework achieves more than 90% accuracy and recall rate,which is comparable to SegNet and FCN 16 Scene Parsing framework.Its real-time applications can be achieved both in X86 and ARM platform.
-
Computer Engineering. 2017, 43(7): 298-302,308. https://doi.org/10.3969/j.issn.1000-3428.2017.07.050In order to support aircraft structure damage analysis and intelligent maintenance decision accurately and efficiently,an improved airframe damage region division method is proposed.According to the features of airframe damage region,mathematical morphology is introduced to repair the image which is divided by gray level entropy.On the basis,the single pixel initial contour extraction method is proposed.Snake model is improved by automatic judgment of stable boundary points and dynamic addition of new boundary points.Airframe damage region division experiments are performed by different types of damage image.Experimental results show that,compared with the classical Snake methods,the proposed method solves the influence of initial contour uncertainty on the results effectively,and avoids disturbance of neighboring regions.The damage region divided by the proposed method is clear and accurate and the operation efficiency is improved obviously.
-
Computer Engineering. 2017, 43(7): 303-308. https://doi.org/10.3969/j.issn.1000-3428.2017.07.051In recent years,the unmanned vehicle has become a hot research direction,and negative obstacle detection is one of the tasks in environmental perception and understanding for unmanned ground vehicles.Therefore a negative obstacle detection algorithm based on single line laser radar and monocular vision is proposed.To compensate for the lack of coverage of single line laser radar,it is necessary to track the detected negative obstacle area in the camera screen,and then the negative obstacle area is further determined according to the tracking results.The experimental results show that the algorithm has more than 96% negative obstacle detection accuracy in a variety of experimental scenarios.It can be applied to mini unmanned ground vehicles effectively.
-
Computer Engineering. 2017, 43(7): 309-315. https://doi.org/10.3969/j.issn.1000-3428.2017.07.052The primary goal of the dredging operation system is dredging production maximization.Safe and stable running of the soil cutting swing process and the slurry pipeline transmission process is required,which is summed up as a problem of linear quadratic optimal tracking control with state inequality constraints.In view of this problem,this paper proposes a transformation method of augmenting dimensions of system state space.This method introduces auxiliary state variables and control variables,transforms the state inequality constraints into equality constraints,and constructs an equivalent dredging system with control security.Simulation results show that the proposed method can increase the slurry concentration.At the same time,the overload operation of the system’s main state for a long time can be limited,so as to effectively enhance the safety and stability of the dredging operation.
-
Computer Engineering. 2017, 43(7): 316-321. https://doi.org/10.3969/j.issn.1000-3428.2017.07.053In the network maintenance support of the command and control system,aiming at the problem of strong subjectivity and low real time property in the process of health assessment,to reduce the influence of human factors,by analyzing the geometric significance of Support Vector Machine(SVM) classification,the problem of network health assessment is transformed into a classification problem.Optimizing Samples-SVM(OS-SVM) is used to optimize the sample points,which can improve the classification speed and classification performance of samples.In the correspondence between quantitative analysis and qualitative description of health status,the cloud model is used to realize the transformation of uncertain information.Experimental results show that the simulation results obtained by the improved command and control network model are in good agreement with the actual situation,and the classification performance is higher.
2024
Volume 50 No 3
Volume 50 No 3
(Started in 1975, Monthly)
Sponsored by：
East-China Institute of Computer Technology Shanghai Computer Society
Edited & Published by：
Editorial Board of Computer Engineering
Chief Editor：Jiang Bo
Executive Chief Editor:
Hu Bing
Tel：021-67092217
E-mail: ecice06@ecict.com.cn
CN: 31-1289/TP
ISSN: 1000-3428
Sponsored by：
East-China Institute of Computer Technology Shanghai Computer Society
Edited & Published by：
Editorial Board of Computer Engineering
Chief Editor：Jiang Bo
Executive Chief Editor:
Hu Bing
Tel：021-67092217
E-mail: ecice06@ecict.com.cn
CN: 31-1289/TP
ISSN: 1000-3428