15 February 2018, Volume 44 Issue 2
-
Select all|
-
Computer Engineering. 2018, 44(2): 1-8.Domanial microblog contains much professional information that show a strong evolution over time.In order to analyze the topics of professional microblog automatically,a domanial microblog topic evolution method based on Hierarchical Dirichlet Process(HDP) model is built.Firstly,domain-related microblog is extracted with the individual user as the unit.Then,accurate extraction of domain-related microblog with distinct temporal features and automatic mining of its topics using domain features and temporal features.At last,the process of domanial topics evolution analysis is constructed.Experimental results show that the method based on the DM-HDP model can show the evolution of the field of microblog,and compared with the methods that based on the LDA and HDP model,it has obvious advantages in terms of content confusion and model complexity.
-
Computer Engineering. 2018, 44(2): 9-16.The traditional similarity measurement algorithm of time series needs human intervention such as time alignment,which has higher time complexity and is bad for following data mining.For the problems above,this paper puts forward similarity measurement algorithm of time series based on coefficient matrix arc differential.It introduces the thought of least-square method in regression analysis,obtains vector basement of time series form attribution by constructing matrix and achieves the continuous sequence curve simultaneously.On this basis,it implements similarity measurement of time series finally by using the relationship of arc differential and curvature radius of the continuous function.Experimental results show that the proposed algorithm has stronger robustness,which can not only achieve similarity measurement in microscopic(lies in distance proximity),but also achieve similarity measurement in macroscopic(lies in same configuration).
-
Computer Engineering. 2018, 44(2): 17-23,30.
Text interval hot word query is based on user-specified query time range,it extracts from the text data hot words.Existing hot words extraction algorithm is generally oriented to mining tasks,which has a high time complexity and is difficult to be directly applied to an online query processing of hot words.Therefore,an online query processing algorithm for text interval hot words is proposed.Using data partitioning and range search technology,the time complexity of extracting hot words is reduced with the same accuracy and space complexity.Experimental results show that compared with the existing mining-oriented algorithms,the running time of the algorithm is reduced by 59.7%,65.1% and 75.5% respectively over the entire time range covered by the three CNN,BBC and NYT datasets,which effectively improves the hot words online query efficiency.
-
Computer Engineering. 2018, 44(2): 24-30.In text data stream,frequent concept drifts result in the poor effective information,thus the accuracy rates of drift detection and stream classification are lower.To address this problem,by introducing Latent Dirichlet Allocation(LDA) model and considering the semantic information of text data stream,this paper proposes a new concept drift detection algorithm.It calculates the semantic similarities of both word and topic feature spaces between adjacent modules,in which the similarity of topics is evaluated by the probability distribution of topic-word.It is considered that concept drifts occur when the similarities are lower in these two spaces.Experimental results show that,compared with DDM,CDRDT,DWCDS,HDDM-W-Test and REDLLA algorithms,the proposed algorithm can improve the performance in the concept drift detection.Especially,it can significantly reduce the missing drifts when concept frequently drifts.
-
Computer Engineering. 2018, 44(2): 31-39,45.When predicting influenced nodes in Temporal Diffusion Network(TDN),it is costly to verify the states of nodes.Besides,it is #P-hard to compute the infection probability of nodes.Aiming at these problems,based on Hop Limited Approximative(HLA) and iterative update algorithm,this paper proposes IPH algorithm.It computes the infection probability of nodes approximately,verifies the state of node that has maximum infection probability in candidate set and updates candidate node set after verification.As infection probability of remain nodes are very close after verifications,this paper futher proposes AIPH algorithm based on Breath First Search(BFS) algorithm to solve this problem.Experimental results show that IPH algorithm and AIPH algorithm have better performance compared with BFS algorithm and Random Work(RW) algorithm.
-
Computer Engineering. 2018, 44(2): 40-45.The master node will easily become a bottleneck in SensorFS due to the centralized sensor clustering algorithm.Besides,it will cost a lot of time when there are massive data.In that way,distributed sensor clustering algorithm and load balancing algorithm based on sensor are put forward.The master node is only responsible for initial scheduling of sensors.Then,the sensor interacts with specific ChunkServer node directly.Inside each ChunkServer node,the sensors are divided into multiple classes by sensor clustering algorithm based on Sensing Dependency Graph(SDG),and these sensor classes can be clustered by master node.Furthermore,taking the different file arriving-rate of each sensor into account,the load balancing is executed based on sensor class.Experimental results show that,in case of massive small data in Hadoop Distributed File System(HDFS),distributed sensor clustering algorithm and load balancing algorithm based on sensor class can effectively improve the read/write performance of the system for massive small-size data.
-
Computer Engineering. 2018, 44(2): 46-50.In order to replacing cache efficiently in Named Data Networking(NDN),a cache replacement strategy combining dynamic popularity and cost cache replacement policy is proposed in this paper,which named DPCP.Each node calculates the Dynamic Popularity and Cost(DPC) value of every cache content in its content store.It replaces cache based on the DPC value of content store’s cache content,so as to keep content with high popularity and request cost.Furthermore,it divides contents into different types based on the DPC value and executes Differentiated Decision Policy(DDP) to choose cache node.Experimental results show that the proposed strategy can achieve higher cache hit ratio and reduce average request hop,compared with classical NDN cache algorithm.
-
Computer Engineering. 2018, 44(2): 51-55.Aiming at the problems in Mixed-Criticality System(MCS) tasks scheduling,such as the low utilization of heterogeneous multi-core,the negative treatment about non-critical tasks,critical tasks cannot migrate in different cores and so on,this paper proposes a novel mixed-criticality tasks scheduling algorithm which is suitable for the heterogeneous multi-core system.In the stage of processor allocation,it allocates the critical tasks to more powerful processors,assigns the mixed-criticality tasks with the heuristic algorithm and takes the maximum residual utilization as the index at the same time.Meanwhile,the recovery queue is introduced to deal with the non-critical tasks that are discarded.The results of simulation show the effectiveness and superiority of the proposed algorithm in improving the acceptance ability of both critical and non-critical tasks.
-
Computer Engineering. 2018, 44(2): 56-60,67.
The default LCE caching strategy of the Information-Center Networking(ICN) which caches content on each node on the return path of the packet generates a large number of redundant copies and cannot make full use of the cache resources.To solve this problem,a caching strategy based on heat and cache replacement rate of node is proposed.It chooses special nodes on packet return path cache content.Considering the network traffic of regional differences and different time difference,it computes heat cache and replacement rate of node periodically,as a measure of whether the content is cached in node.Experimental results show that,compared with LCE and CLFM strategy,the proposed strategy can reduce average number of hops and server hit ratio,meanwhile getting higher caching benefit.
-
Computer Engineering. 2018, 44(2): 61-67.In order to meet the higher real-time requirements of event-driven wireless Cyber-physical System(CPS),this paper proposes a real-time message parallel scheduling method.The method takes the deadline of message as the main factor.It considers the residual energy of the node and the average delay of the network.Based on these factors,the base station selects the optimal receiving node for each sending node of the message,and chooses an appropriate transmission path.At the same time,using the graph coloring theory and the tabu search algorithm,the message graph corresponding to the required transmission message in each time slot is colored multiple conditions.The messages corresponding to the vertices of the same color can be sent in parallel so as to minimize the number and maximize the parallel transmission degree.Theoretical analysis and simulation results show that this method has a lower deadline loss rate,which can reduce the end-to-end delay and reduce energy consumption.
-
Computer Engineering. 2018, 44(2): 68-74,83.Aiming at the problem of deep neural network speeding up training on distributed multi-machine and multi-GPU,this paper proposes an implementation method of remote multi-GPUs calls based on virtualization.The distributed GPU clusters deployed by remote GPU calls improve the traditional one-to-one virtualization technology and change the location of the deep neural network for parameter exchange during distributed multi-GPU training,achieve the compatibility between the two.The method utilizes the remote GPU resources in a distributed environment to speed up the training of deep neural networks,and reaches the unification of CUDA programming modes of single GPU and multi-GPU.Taking handwritten numeral recognition as an example,experiments are carried out on the parallel training of multi-GPU and multi-GPU data in the deep network of general network environment,results verify the effectiveness and feasibility of the method.
-
Computer Engineering. 2018, 44(2): 75-78.Faster-than-Nyquist (FTN) signaling system can increase data transmission rate effectively.However,it induces unavoidable Inter-symbol Interference(ISI) effects at the receiver and leads to higher receiver complexity.Against this problem,people put forward Iterative Block Decision Feedback Equalizer(IBDFE) and Low Complexity Iterative Block Decision Feedback Equalizer(LC-IBDFE),and these two algorithms can greatly reduce computational complexity of FTN.Simulation results show that BER performance of these two algorithms are almost the same in AWGN channel,and LC-IBDFE can achieve much lower computational complexity.In order to promote the practical application of FTN,this paper expands IBDFE and LC-IBDFE to frequency-selective fading channels.Simulation results show that these two algorithms can still detect the FTN signal when increasing the length of PN sequence appropriately.Meanwhile,the rate of effective information decreases.
-
Computer Engineering. 2018, 44(2): 79-83.In Underwater Wireless Sensor Network(UWSN),limited resources (bandwidth and energy) of the sensor nodes hinder the data transmitting from a mobile sensor to any sonobuoy.Therefore,a novel opportunistic routing for UWSN is proposed in this paper.It introduces depth threshold to select the reliable and lower depth forwarder node which improves goodput and reduces number of hops.Whenever a sensor node needs to send a packet,it uses the depth to determine the candidate forwarding set,then computes the path metric of each node in candidate forwarding set,which can reflect the reliability of link,and order by path metric.Finally,the redundant packet is suppressed by timer.Simulation results show that this protocol has better energy utilization and network goodput compared with pressure based sensing protocol.
-
Computer Engineering. 2018, 44(2): 84-87,91.Traditional (k+2,k) Minimum Storage Regeneration Codes(MSR) have a high repair bandwidth in the event of two system nodes fail.A safe,efficient,multi-fault tolerant coding scheme is proposed for this reason.The upper and lower parts of code are both formed MSR codes structure through the introduction of four backup check nodes.Simulation results show that compared with the existing (k+2, k) MSR code scheme,this scheme can greatly reduce the repair bandwidth when two node fails.
-
Computer Engineering. 2018, 44(2): 88-91.In order to alleviate the rapid growth of network data brought about energy consumption,in response to the request of energy saving and emission reduction,improving energy efficiency,presents an energy-efficient routing algorithm.The algorithm uses a complex gradient network constructed with a scale-free network as a substrate network,a node “potential” defined by the betweenness of neighbor nodes,and the gradient-driven transmission strategy for packet forwarding and transmission is constructed with the gradient of the node potential.Simulation results show that this paper compares the shortest path routing algorithm,in the larger network data generation,the algorithm in this paper can bypass nodes with large betweenness to avoid congestion,thus effectively reducing network energy consumption,reducing packet forwarding time,achieving the purpose of network energy efficiency optimization.
-
Computer Engineering. 2018, 44(2): 92-97,102.In the model of Multi-way Relay Channel (MRC) with multi-input multi-output,the communication process is inevitably affected by the interference of streams,resulting in high bit error rate and poor communication quality.In this paper,a coding scheme based on nested lattice is proposed.This coding scheme encodes the messages via the special geometrical mechanism of lattice,and improves the capacity with the help of nestification of different layer of lattices.The scheme can solve the communication problems and improve the performance combined with the user-side pre-coding.Simulation results show that the proposed scheme is better than the traditional Decode and Forward(DF) and Amplify and Forward(AF) scheme,and good link channel capacity can be reached.
-
Computer Engineering. 2018, 44(2): 98-102.According to the characteristics of the current communication source identification method needs not only the higher number of samples,and identification efficiency is low,recognition rate drop,this paper proposes a Softmax regression was applied to cyclic spectrum signal classification method.The method based on cyclic spectrum density features of communication signals sample set,filtering feature samples by Principle component analysis dimensionality reduction algorithm,finally using Softmax return classification recognizer classify characteristics of samples.Experimental results show that compared with the traditional neural network algorithm,the method can achieve an efficient identification of communication sources,and the identification of a relatively short time.
-
Joint Power and Bandwidth Allocation Algorithm with Inter-beam Interference for Multi-beam SatelliteComputer Engineering. 2018, 44(2): 103-106,113.In consideration of the inter-beam interference in multi-beam satellite communication systems,a model of a multi-beam satellite system with inter-beam is set up by introducing with interference factor.Then,a joint power bandwidth allocation algorithm is proposed.The resource allocation problem is divided into three sub-problems:power allocation,bandwidth allocation and renewal of dual variables.During the algorithm,firstly it updates the dual variables by using the sub-gradient algorithm after power and bandwidth allocation for the first time.Then it redistributes the power and bandwidth until the dual variables is convergent.By using the joint algorithm proposed,it can guarantee the result is the optimal solution.Simulation results show the performance of the system is improved compared with the existing algorithms.
-
Computer Engineering. 2018, 44(2): 107-113.As to the problem that the resource allocation technology is not adaptive in the hybrid network of cellular and Device-to-Device(D2D) which causes the waste of resources,a resource allocation scheme is proposed which can be adjusted adaptively according to the actual network environment,and a two-stage resource allocation algorithm is designed to solve it.Both the number of users using every resource block and the number of resource blocks that D2D users can use are adjusted adaptively according to the interference between users in the first stage,and the improved PSO algorithm is used to allocate power which maximizes the throughput in the second stage.The simulation results show that the proposed algorithm is near optimal algorithm.Besides,the throughput of the system and the access rate of D2D users are significantly superior to the fixed allocation algorithm.
-
Computer Engineering. 2018, 44(2): 114-118,123.In order to solve the problem of secure communication in Vehicle Ad Hoc Network(VANET),an abnormal node detection mechanism based on Greenshield model is proposed.Combined with the characteristics of vehicle ad hoc networks,Greenshield model is constructed to calculate the vehicle speed,vehicle density and traffic flow parameters.On this basis,based on the difference between the vehicle traffic calculated by the vehicle wireless communication devices and the traffic flow calculated by other vehicles received,the location of possible abnormal nodes may be initially located.The u test method in the hypothesis test is adopted to determine whether to accept the received data and to infer whether the node is abnormal.Simulation results show that the real index of abnormal node detection using this mechanism is high,the index of false positive rate is low,and the abnormal node in the vehicle ad hoc network can be effectively detected.
-
Computer Engineering. 2018, 44(2): 119-123.In order to improve the performance of transmission control protocol on ship ad-hoc network,the problem of congestion control and exposed terminal in high load network is analyzed,the congestion control algorithm is improved by starting a timer,the semi transmission control protocol named Semi-TCP-RTS-V2 is proposed,and the performance of TCP-Lite,Semi-TCP-RTS,Semi-TCP-RTS-V2 is analyzed in the Exata simulation platform,the results show that the transmission performance of Semi-TCP-RTS-V2 in the ship ad-hoc network is improved.
-
Computer Engineering. 2018, 44(2): 124-128,134.The importance of complex network nodes ordering is an important part of complex network research.Based on the k- core algorithm,weighted value is introduced to redefine the value of node k for weighted network.In order to quantify the influence of weight on the importance of network nodes,a new improved k-core algorithm for weighted network is proposed and the influence of the change of the balance coefficients value on the algorithm is analyzed.Simulation results show that the proposed algorithm is superior to k-core,and has the characteristics of adjusting the balance coefficient to adapt to different weighted networks,it is suitable for the importance evaluation of weighted network nodes.
-
Computer Engineering. 2018, 44(2): 129-134.The slow target detection in the background of strong clutter has such problems as low Doppler frequency shift,clutter interference,lack of robustness,feature extraction difficulties and inadequate information utilization.Therefore,a target detection method of wideband signal based on deep self-coding network is proposed.The echo information is analyzed by using time-frequency transform,and the deep self-coding network algorithm is used to extract the target deep abstract information in the time-frequency domain for target detection to accurately sense the environmental change.Simulation results show that compared with traditional machine learning such as Support Vector Machine(SVM),Extreme Learning Machine(ELM) and Back Propagation Neural Network(BPNN),the proposed method can effectively detect environmental changes and has high robustness and detection performance.
-
Computer Engineering. 2018, 44(2): 135-140,146.Signal routing is a key technology of signal-oriented autotest system.The existing signal routing technology generally only can establish a signal propagation path between a single pair of pins and can not satisfy the condition that multiple excitation signals are routed at the same time.therefore,this paper discuss the multipath signal routing algorithm of automatic test system satisfying certain conditions.The hardware environment of automatic test system is abstracted as flow network,on the basis of this,which the propagation path of multi path signal is constructed by solving the maximum flow.To improve the quality of signal propagation,the concept of “propagation resistance” of the switch matrix is defined and the signal propagation path is optimized by the minimum cost maximum flow algorithm.Experimental results show that the algorithm can automatically generate routing strategies for multiple signals.
-
Computer Engineering. 2018, 44(2): 141-146.Sparse Fast Fourier Transform(sFFT) requires the signal to have the sparsity of the Fourier domain as prior information,but the sparsity is usually unknown,which limits the application of the algorithm to a certain extent.therefore,a new sparse Fourier transform algorithm is proposed.The energy is detected in the downsampling domain to get the initial value of the sparsity.The accuracy of the sparsity estimation is increased by increasing the downsampling dimension so as to estimate approximatively the sparsity,and the threshold is set to eliminate the redundant information to obtain the better result.Experimental results show that the performance of the algorithm is superior to Faster Fourier Transform in the West(FFTW) when the signal sizes is greater than 219 or the sparsity is less than 900,and the algorithm has strong robustness.
-
Computer Engineering. 2018, 44(2): 147-150,162.The existing Frequency Hopping(FH) sequence can not be directly applied to a dynamic spectrum anti-interference system due to poor statistical performance and difficulty in real-time changes of communication parameters.Therefore,a FH sequence suitable for dynamic spectrum anti-jamming system is proposed.Based on m sequence,a hopping base sequence is generated by discontinuous model.According to the idea of stochastic translational substitution,a pseudo-random mapping of base sequences is proposed,and a dynamic wide interval frequency hopping sequence congtruction method with frequency number and FH interval is qiven according to the changed communication environment.Simulation results show that the dynamic wide-gap hopping sequence has better performance in terms of uniformity,randomness,Hamming correlation,average hopping gap and so on,compared with the frequency-adaptive FH sequence based on fixed parameters.
-
Computer Engineering. 2018, 44(2): 151-157.Aiming at the problem that the chaos based image encryption algorithm combined with DNA coding can be broken caused by incorrect use of DNA coding rules and calculation methods,a new image encryption algorithm combined Intertwining Logistic mapping and dynamic DNA coding is proposed in this paper.This algorithm includes a scrambling phase and a diffusion phase.In the scrambling phase,the rows and columns of each channel of the color image are scrambled via scrambling sequences constructed by the simulated annealing algorithm.In the diffusion phase,the DNA coding rules and calculation rules of each pixel are determined according to the random sequence values,and then the DNA encryption is carried out.Experimental results show that this algorithm can resist against common attacks and have an excellent performance.
-
Computer Engineering. 2018, 44(2): 158-162.In order to enhance the ability of the stream cipher algorithm for Leak Extraction(LEX) in resisting slide attack,16 bytes from the intermediate variable are extracted at each key expansion of Advanced Encryption Standard(AES),and they are used as the secret key K in the next AES.The key system of LEX is improved.On this basis,the safety and operating speed of the improved LEX algorithm is analyzed.The simulation based on C++ is also conducted to test the key stream randomness of the improved LEX algorithm.The results show that the improved algorithm can resist slide attack with the same computing speed and key stream randomness in the LEX algorithm.
-
Computer Engineering. 2018, 44(2): 163-170,176.In the rapid development of mobile Internet,the security of Android devices becomes more and more prominent,bringing many security risks to mobile Internet users.In this paper,the authors introduce related researches on Android security protection in recent years,point out its advantages and disadvantages,and put forward some improvements.By comparing and analyzing the existing work and related security protection technologies,the challenges and opportunities of Android security protection are given,and the broad prospects of the field of Android security protection are prospected.
-
Computer Engineering. 2018, 44(2): 171-176.
The current research of access control mechanism is clearly distinguished between the subject and object,and user’s permission is relatively fixed,but in practical applications,user’s permission needs to be adjusted according to the change of the service environment.This paper presents an environmental attribute-based access control model,designed and implemented a prototype system under the Linux system.The system adds environmental attribute factors into the access control decision process,adjusts user’s permission dynamically.Based on small system overhead,it can enhance the system security and availability without affecting the operating efficiency of the Linux platform.
-
Computer Engineering. 2018, 44(2): 177-181.Because of the broadcast characteristics of wirless network,in the process of signal transmission,information leakage often occurs,so that there are hidden dangers of information security.In view of this,an Orthogonal Frequency Division Multiplexing(OFDM) system is proposed to generate transmission signals.In combination with the diversity of wireless channels and the inherent randomness of wireless multipath channels,an envelope and phase are extracted from the channel as parameters to generate wireless channel based on wireless multipath channel parameters Key scheme.Experimental results show that this scheme increases the key consistency,reduces the key inconsistency rate,and increases the security of the information by the channel characteristics of key generation.
-
Computer Engineering. 2018, 44(2): 182-186.In the topic of distinguishment of orthogonal states in dd bipartite system,with the way of judging dimension of there corresponding orthogonal complement,and then determine whether the orthogonal complement space contains a maximally Abelian subspace,at last propose a new algorithm to distinguish states based on one-way LOCC.For generalized orthogonal Bell states in 44 bipartite system,by studying their distinguishment,this paper propose a faster algorithm based on one way LOCC.Experimental results show that the algorithm in ddbipartite system can be applies in distinguishing any orthogonal bipartite states.What’s more,the algorithm in 44 bipartite system is less complicated.
-
Computer Engineering. 2018, 44(2): 187-192.In order to improve the security of Android application software and increase the difficulty of the attacker without affecting the running efficiency of APP as the target,an APP security reinforcement system is designed on the basis of the automatic identification of APP documents.The system identifies the key words by automatic identification of the APP documents.In addition,it determines the security requirements of APP according to the key words,and then gives the corresponding security combination and reinforcement scheme by the security demands to realize the concrete safety reinforcement.The analysis results show that the system can increase the security of APP with the appropriate reinforcement scheme,and protects the legitimate interests of users and developers effectively.
-
Computer Engineering. 2018, 44(2): 193-196. https://doi.org/10.3969/j.issn.1000-3428.2018.02.034In order to improve the operating efficiency and the security of encryption and decryption of attribute encryption system,the theory of lattice instead of bilinear pairings is used to reduce the computational complexity of encryption and decryption process.According to the scheme of ciphertext policy attribute encryption from lattices and the scheme of dynamic multi-attribute encryption,a multiple encryption scheme is proposed based on lattices.This scheme has the capability to encrypt multiple messages at the same time and improve the operating efficiency of the system,and it can also against quantum attack.The correctness of the scheme is strictly derived,and the security of the scheme is reduced to Learning With Errors(LWE) hard problem using provable security.The analysis result shows that the proposed scheme is correct and feasibile.
-
Computer Engineering. 2018, 44(2): 197-202.In order to improve the accuracy of semantic correlation of short text,this paper uses Wikipedia as an external semantic knowledge base,and combines with structure features of Wikipedia,such as typing architecture of Wikipedia,link structure between abstracts and pages,and redirect disambiguation pages,and puts forward the calculation algorithm for the correlation between words.On this basis,it also puts forward the sentence correlation calculation method combined with word order structure and weight of subject words.Experimental results show that in terms of word correlation calculation,Spearman parameter of the method in this paper is 2.8% higher than that of the text correlation calculation methods,and the accuracy of sentence correlation is up to 73.3%.
-
Computer Engineering. 2018, 44(2): 203-209,219.Opinion leader is active information promulgator in the network community and the guide of information transmission.The assessment for opinion leaders’ influence is an important content of social network analysis.To improve the considerations about the analysis of user dynamic behavior and the influence of dynamic content,and reflect the real situation objectively,a discovery algorithm of opinion leaders based on user’s influence and PageRank is proposed.It combines user’s own influence,the influence degree of user dynamic behaviors and the real impact of dynamic content brought by the dynamic behaviors.Experimental results on the large-scale data collected from Zhihu network community demonstrate that the algorithm is more reasonable and can effectively improve the recognition accuracy of the network community opinion leaders.
-
Computer Engineering. 2018, 44(2): 210-219.In order to classify the emotion for users expressions and comments on social networks,this paper presents a sentiment classification method which combines Convolutional Neural Network(CNN) and multi-feature fusion.It designs corpus features and lexicon features according to the characteristics of Twitter texts and semantic lexicons,uses the convolution neural network for the word vector of Twitter text to get the depth word vector features,combines the above three features to construct the feature fusion and uses One-Versus-One SVM to obtain the sentiment polarity classification and discrimination.Experimental results on SemEval corpus show this method performs a good result and the multi-feature fusion can efficiently improve the accuracy of sentiment classification.
-
Computer Engineering. 2018, 44(2): 220-225.A unsupervised approach to normalize the irregular Uygur words in the spoken text to normal words in the formal text.Based on neural network,Uygur words are mapped to a low dimensional vector space by using a large corpus.The irregular words in vector space are clustered.A greedy decoder is introduced to normalize the unformal words and to resample iterations,so as to normalize the unformal words which have not been successfully normalized before.Experiment results show that using this approach to pre-edit the text to be translated by Uyghur-Chinese machine translation,the quality of the generated translation is significantly improved.This method provides a pretreatment process to spoken text and machine translation system,which can effectively improve the system performance of machine translation in the absence of bilingual parallel corpus of spoken.
-
Collaborative Filtering Recommendation Algorithm Based on Representation Learning of Knowledge GraphComputer Engineering. 2018, 44(2): 226-232,263.To solve the problem that collaborative filtering algorithm only uses the items-users rating matrix and does not consider semantic,a collaborative filtering recommendation algorithm is presented.Using the knowledge map to represent the learning method,this method embeds the existing semantic data into a low-dimensional semantic space.It integrates the semantic information of items into the collaborative filtering recommendation by calculating the semantic similarity between items.The shortcoming of collaborative filtering algorithm which does not consider the semantic information of items is overcome,and therefore the effect of collaborative filtering recommendation is improved on the semantic level.Experimental results show that the proposed algorithm can get higher values on precision,recall and F-measure for collaborative filtering recommendation.
-
Computer Engineering. 2018, 44(2): 233-237,270.Currently,most word vector models can build only one vector for a single word.Due to word’s polysemy,it is incorrect to use one vector representing a same word under different context.This paper proposes a new word vector model.It uses latent dirichlet distribution and neural networks to train words to obtain word vectors and corresponding topic vectors.And then it applies linear transformations on them to build the final word vectors.Experimental results show that the accuracy of proposed model is high compared with current multi-vector models.
-
Computer Engineering. 2018, 44(2): 238-243,276.This paper proposes an Ant Colony Optimization(ACO) algorithm for gradually adaptation and adjustment of pheromone concentration to improve the learning efficiency of learners in view of the characteristics of micro-learning,which optimizes the recommendation of micro-learning paths.In the whole process of micro-learning,the learning state of the learner is acquired through the interaction of the learner and the system,and the strategy of the learning path is adjusted according to the learning state.The learning path is adjusted in the granularity of the learning unit,thus achieving the goal of catching the individual needs of the learners and helping the learners improve their learning efficiency.
-
Computer Engineering. 2018, 44(2): 244-250,281.The velocity updating mechanism of Particle Swarm Optimization(PSO) algorithm is Proportion-Integral(PI) control strategy.On account of the inherent integral term,the system is prone to oscillation,which leads to the low search speed.Therefore,according to the characteristics of Proportional-Integral-Derivative(PID) control,this paper proposes a fast PSO algorithm.The differential control is added in the standard PSO algorithm and its improved algorithm to overcome the oscillation,improve the convergence speed,and increase the stability of the search process in the meantime.Simulation results show that,compared with the standard PSO algorithm and the full information PSO algorithm,the proposed algorithm improves the search speed and has high computing efficiency while ensuring the optimization accuracy and reliability.
-
Computer Engineering. 2018, 44(2): 251-256.Traditional face image retrieval technology has lower retrieval efficiency when processing large-scale image data.Aiming at this problem,a face retrieval system based on Bag-of-Visual-Words(BoVW) model and Spark distributed platform is constructed in this paper.A local block partition method is proposed according to the spatial distribution of a face image,so as to reduce the number of visual features and enhance parallelism.By combining the Speed-up Robust Feature(SURF) local features and Histogram of Oriented Gradient(HOG) block features,a similarity algorithm of candidate images is designed to improve the retrieval accuracy.Experimental results show that the efficiency of the index construction and image retrieval in the proposed system are higher than those of the retrieval system based on Hadoop.The proposed system also has good scalability and concurrency under the massive image data scene.
-
Computer Engineering. 2018, 44(2): 257-263.The action recognition method of step learning adopts the idea of curriculum learning,ignores the influence of different angles movement characteristics on the course,and can not achieve satisfactory results for the classification of two dimensional video complex action recognition.In order to solve the above problem,an algorithm for Multi-view Self-Paced Learning(MSPL) is proposed.It selects five views and extracts their features (Trajectory,HOG,HOF,MBHx and MBHy),and then learns curriculums under each view by Self-Paced Learning(SPL),fuses curriculums by means of Linear Programming Boosting(LPBoost),and learns a comprehensive curriculum that is more suitable for solving the problem of multi class complex action recognition at last.Experimental results show that compared with SPL and multi-view Support Vector Machine(SVM),the proposed algorithm improves the efficiency and accuracy of multi-class complex action recognition,and has higher operational and wider application prospects.
-
Computer Engineering. 2018, 44(2): 264-270.A Compression Tracking(CT) algorithm is proposed to automatically adjust the learning rate of feature distribution,which is based on the problem that the fixed learning rate is used to update the feature distribution of the tracking algorithm,which is easily affected by the occlusion and the robustness is low.Compressed domain feature samples are obtained by the compressive sensing theory,calculate the distribution characteristics of various compression characteristics in the positive class and negative class,use the distribution of overlap between the two frames combines with adaptive threshold update distribution.Target tracking is achieved by sample classification.At the same time,the algorithm makes use of the improved SIFT features of adjacent two frames to solve the target scale change,and realize the tracking window with the change of the target in real time.Experimental results show that the proposed algorithm can effectively resist the interference of tracking,such as occlusion,ray and scale.It has higher accuracy,robustness and real-time performance.
-
Computer Engineering. 2018, 44(2): 271-276.Aiming at the problem of poor face recognition and low recognition rate in complex environment,the algorithm of extracting facial features by sparse learning and significance theory is proposed in this paper.Through the sparse coding theory to simulate the human visual perception mechanism,constructing multi-scale multi-directional filter to extract image contour feature by using the basis function,the feature is subjected to Local Binary Pattern(LBP) filtering to highlight the face local detail texture feature.According to the visual attention mechanism,the salient features of the treated features are constructed and the contribution of important features to face recognition is enhanced.It uses the LFW,YALE standard library and home-made video frame image library for testing to obtain a higher recognition rate.Experimental results show that the proposed method is superior to the traditional feature extraction method.The obtained facial features are more representative and it has strong robustness in the complex environment.
-
Computer Engineering. 2018, 44(2): 277-281.The understanding of 3D meshes has become an important problem to be solved in many geometric processing problems,and the mesh segmentation is the key step,therefore a novel mesh segmentation method based on Ant Colony Optimization(ACO) is proposed in this paper.Each mesh of the mesh is regarded as an ant,and the label of each grid is updated by ACO iterations.With the iteration of the ACO,the labels of the seed points spread outward and the updating of the labels is performed by the ACO mechanism while satisfying the segmentation criterion,until the iteration criteria are met.After the ACO is completed,the regions are merged and the smaller regions are merged into larger regions.The proposed method is implemented and a part of the Princeton grid data set is tested.The experimental results show that it can achieve higher accuracy compared with the graph cut segmentation method.
-
Computer Engineering. 2018, 44(2): 282-286.The traditional Convolutional Neural Network(CNN) is difficult to obtain high recognition rate when the feature of the input image is not obvious.To solve this problem,this paper conatrusts Convolutional Neural Network model Based on Local Feature(CNN-LF),which is built on CNN by adding Local Feature Extration(LFE) layer and Probability Importance Synthesis(PIS) layer.It firstly recognizes the local feature of the input image and gets the information about classification probability,then makes comprehensive analysis on the information of all the local images to get the final result.In the experiments of handwritten digits recognition,CNN-LF gets higher recognition rate compared with traditional network model,especially when recognizing images which have fuzzy features.
-
Computer Engineering. 2018, 44(2): 287-293.Taking the regional load as the research object,the differences of volatility and periodicity between regional load and grid load are analyzed.According to the characteristics of weak temporal correlation in the process of regional load change,a kind of ultra-short term regional load prediction method based on pyramid model is proposed.The gray relational analysis method is used to extract the objective characteristics of strong correlation with load change.The Adaptive Boosting Random Weighted Network (Ada-RWN) model is established to enhance the ability of load trend characteristics learning and optimal solution efficiency.A hierarchical pyramid model structure,which uses the method of rolling out to improve the adaptability of the forecasting model to the regional load characteristics,is designed to reduce the influence of the regional load variation trend on the ultrashort prediction accuracy.Simulation results show that the proposed method can follow the trend of regional load change and has high prediction accuracy and stability.
-
Computer Engineering. 2018, 44(2): 294-297.According to the characteristics of command and control network topology and cascading failures,this paper constructs the cascading failure model in command and control network.The model takes network robustness as the invulnerability measure.The nonlinear load-capacity model and the load capacity redistribution rule are used to simulate the cascaded failure process.By simulation analysis of the influence of load parameter,capacity parameter,evolution step size on cascading failure invulnerability in command and control network,it provides a certain reference for study of cascading failure behavior in command and control network.
-
Computer Engineering. 2018, 44(2): 298-303,309.For some manufacturing exists complex industrial processes with severe nonlinearity,named pH neutralization process,a new process T-S fuzzy modelling method via objective clustering idea,combined with Gustafson-Kessel(GK) clustering,is proposed.According to the satisfactory degree of modeling performance indexes,the premise structures and their parameters are identified through the iterated fuzzy clustering.Simulation results show that compared with the traditional fuzzy clustering,the proposed method does not rely on the prior-knowledge or defined fuzzy membership function,has a streamlined structure and better approximation performance,and is robust for the existing noise in data.
-
Computer Engineering. 2018, 44(2): 304-309.The change of characteristic parameters can make the browser fingerprint change,resulting in a user produce multiple fingerprints,and the static matching method of fingerprint cannot identify all of the return visit users.Aiming at this problem,this paper proposes a recognition algorithm of the return visitors.It defines the difference degree calculation method of fingerprint,and according to the difference degree to determine whether the fingerprint is the update one of the fingerprint in the database,and then judge whether the user is the return visit user.Experimental results show that the algorithm can effectively identify the return visitors.
-
Computer Engineering. 2018, 44(2): 310-315.In target tracking system,because the sensor has different preprocessing time and sampling rate,and the inherent random communication delay of the channel,the phenomenon of random arrival of the fusion center may appear in the sensor data,that is,the problem of disorder measurement.In the process of system operation,there are usually a number of Out-of-Sequence Measurement(OOSM) appearing in succession or at the same time.Aiming at this problem,classifying multiple disorder measurements,a filtering algorithm on arbitrary-step-lag out-of-sequence measurements based on selective fusion is put forward.The algorithm uses log likelihood ratio hypothesis test to choose the out-of-sequence measurements.Then,according to the optimal OOSM filtering process,the state estimation and the covariance matrix with the information filtering method blended in equivalent measurement within the forward prediction framework is updated.Simulation results verify the precision and effectiveness of the proposed algorithm.
-
Computer Engineering. 2018, 44(2): 316-321.In Speaker Change Detection(SCD) of rapid conversion condition with short speech segment,speaker models training from deficient speech frames of a speaker are not rubust enough,and SCD performance is less satisfied.Therefore,a new SCD method based on Computational Auditory Scene Analysis(CASA) is proposed.The speech signal is decomposed into a number of narrow sub-band signals owing to the auditory processing mechamism of human ears.Accurate voiced speech and unvoiced speech boundaries are obtained,voice sub-segments is spliced from scattered voice and unvoiced sub-segments.Speaker change points are determined between the speaker voice sub-segments by Bayesian Information Criterion(BIC),pitch features extracted from voiced portion are used to verify region.Experimental results show that Equal Error Rate(EER) of SCD can be reduced to 23.2%,which corresponding to 70% of the F1-value,in the rapid conversion situation of average 1.34 s speech sub-segment.
2024
Volume 50 No 4
Volume 50 No 4
(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