- EI核心期刊
A transfer learning based text-image feature mapping algorithm under event constraint is proposed. Firstly, the documents of each event are modeled by the latent dirichlet allocation, in which the most discriminating feature is obtained by computing the information gain of each topic. Secondly, the images of the corresponding event are modeled through the bag-of-visual-word model and the nave bayes approach. Finally, the feature distributions of the target images are approximated by utilizing the feature distributions of the text data and the text-image co-occurrence data within the same event. Experiment is conducted on a dataset containing 15 categories of events. The effectiveness of the proposed feature mapping algorithm is shown.
Digital images can readily be blurred in various degrees and contaminated by noise during the process of the acquisition and transmission. This inevitably leads to considerable difficulties for the subsequent processing and application of these degraded images. An algorithm that combines an alternating minimization method with a new bisection technique for denoising and within-channel/cross-channel deblurring of the color images is proposed in a total variation energy minimization framework. Experiment demonstrates that, due to the fast convergence of the bisection technique employed, the proposed algorithm is able to quickly determine optimal value of the regularization parameter, thus improving the algorithm’s denoising and deblurring performance as well as adaptation ability.
In order to improve the performance of inter-layer intra prediction (ILIP) in scalable video coding, a Wiener filter based block level adaptive ILIP method is proposed. Firstly, the frame to be coded is divided into a number of blocks. Afterwards, the Wiener filter coefficients are obtained by a frame-by-frame training procedure, in which both of the enhancement layer (EL) signal and the corresponding up-sampled base layer (BL) signal are utilized. Finally, the achieved Wiener filter coefficients are applied to optimize the up-sampled BL signal, and a block level filter control is introduced to reduce prediction error energy. Experiments show that, compared with the traditional ILIP method, the proposed method can get more accurate prediction of the EL signal, resulting in up to 14.45% encoding bit rate reduction.
Chinese characters should match certain rules of projection when photoed by digital camera. Meanwhile that will bring projection deviations. In order to detect the information authenticity of the Chinese characters in the digital image, a new estimation algorithm for Chinese characters projection deviation is studied. Based on camera calibration, a new estimation of projection deviation is proposed. And based on Chinese characters’ flat square feature, using Chinese characters model, instead of shooting Chinese characters into the real world, can be coordinated, a set of corresponding point coordinates is extracted, and then the corresponding affine matrix of the image’s homography matrix is obtained. Thus, the deviation value of the reconstruction Chinese character is calculated. If the value obtained is greater than threshold established by experiments before, the Chinese characters in this digital image will be manipulated. Experiments show that this method can detect the rewritten or replaced Chinese characters in the digital image.
A new chaotic communication scheme named orthogonal differential chaos shift keying (ODCSK) is presented for ultra wide bandwidth (UWB) application. The proposed scheme can overcome the disadvantages existing in code-shifted differential chaos shift keying (CS-DCSK), demonstrating desirable features of higher attainable data transmission rate, shows better communication privacy and higher reliability. Analysis and simulation reveal that ODCSK shows better bit error rate performance in comparison to CS-DCSK with certain signal noise ratio levels and spreading factors.
Aiming at shortage of index identified and membership function selected in existing evaluation, a new method of fuzzy evaluation for design scheme base on information axiom is proposed. First, a zigzag mapping model is designed between target area and index area following the principle of axioms design and a comprehensive evaluation index system is established for product design scheme. Second, according to the different relations between index and design range, the evaluation indexes are divided into different types whose membership grade can be described by the normal fuzzy distribution curve. The calculation method is given for the corresponding information of index and can be applied in the scheme evaluation of the mechanical parking system. Last, the process of fuzzy evaluation is modeled based on information axiom.
Based on a pair of Hadamard matrices of the same size, a new construction of mutually orthogonal binary sequence sets with zero correlation zone (ZCZ) is presented by interleaving recursion. The sequence sets proposed can achieve theoretical bound of binary ZCZ sequence sets. By interleaving recursion, ZCZ sequence sets can be constructed from Hadamard sequences weighted by the coefficient matrices, satisfying mutually orthogonality, in which the number of sequences doubles the size of Hadamard matrix. The construction results illustrate that the proposed method improves the number of mutually orthogonal binary sequence sets with ZCZ and obtains more new sequence sets, which are used in quasi-synchronous code division multiple access system and implemented conveniently in hardware, by means of different choices of the coefficient matrices.
A kind of minimum storage regenerating (MSR) codes is constructed for exact-repairing nodes in the distributed storage systems. The codes based on cyclic vectors based are belonged to the maximum distance separable (MDS) erasure codes. Both the system and the redundancy data are saved in each node. All the calculation of the encoding and the decoding are in the Galois field 2. The storage system could exactly repair any single node’s error with the minimum bandwidth in theory. The system could provide a maximum bandwidth in theory for the single user. Compared with other MSR codes, those codes, with minimum bandwidth regenerating repair codes, have more straightforward structure and less decoding operation, and can provide the most available bandwidth for the single user.
Signal power can be enhanced or decreased due to the interference of both none line of sight and path loss in the indoor environment, which causes a bias property of both positive and negative in the received signal strength (RSS) measurement error. A method for improving RSS positioning is proposed. The intersections of positioning annuluses are used in the algorithm of weighted centroid to estimate the position of the mobile stations. Simulation shows that this algorithm can improve the location accuracy in the indoor environment.
A new localization algorithm named connectivity differences distance vector-hop algorithm (CDA DV-Hop) is proposed based on connectivity differences of sensor nodes. First, by estimating the reversing positions of the anchor nodes, the average hop of distance of anchor nodes is refined and the accumulated errors are reduced. Then, according to the characteristic of the similar connectivity in proximate position, limitations of the maximum and minimum of different locations are given under the same unknown node, thus limiting those positions that are beyond the sensing area and calculating the connectivity differences. Finally, the ultimate estimated position is chosen from the candidate positions in which the connectivity difference value is the smallest. Experiments show that, under the same conditions, the proposed CDA DV-Hop algorithm achieves a better precision compared with existing distance vector-hop (DV-Hop) algorithms.
On the base of data glove, a virtual gesture interaction system is put up that realizes the map from mankind hand to virtual hand. The model of space points location in binocular vision is modified and a new model based on two vertically placed cameras is proposed, in which position and posture tracking of moving target by binocular vision are achieved. By modifying camshaft algorithm, the accuracy and automaticity of moving target tracking system are improved, thereafter, the problem of capturing space posture of flexible object is solved. By comparing the pose and position information of virtual hand with image, the algorithm is testified and approved to be applicable.
In order to get the format of unknown protocols, a length semantic constraints based packet format mining method is proposed based on length semantic constraints. First, multiple sequence alignment method is applied to partition a packet into segments. Then, a length identification algorithm is utilized to scan the segments separately to infer length fields and corresponding referred field(s). At last, the format (hierarchy structure) of the packets is obtained. Experiments demonstrate the effectiveness of this method: the false negative rates of length fields for GetNextRequest and GetResponse of simple network management protocol version 1 are both 9.1%, and the false positive rates are 16.7% and 23.1%. The packet hierarchy is also obtained, approximately consistent with protocol format specification.
A new dynamic spectrum access technology based on weighted multi-feature model clustering is presented to improve transmission quality for different users for cognitive wireless network.Once the weighted multi-feature model of channels is extracted, these channels suitable for the cognitive users are clustered into different spectrum pools with a multi-objective clustering algorithm.These channels in white pooling are priority to be accessed for cognitive users according to their services.Multiple feature combinations and dynamic spectrum simulations are completed to verify the efficiency of the proposed method.Analysis and simulation show that the channels can be classified according to performance more efficiently, and the access efficiency of channels can be improved.
In order to break the limitations of spectrum sparse to compressed sensing technology, a pilot detection scheme is proposed based on compressed sensing in non-sparse wideband spectrum sensing. Under the premise of that the interference to authorized users does not exceed the specified threshold, the scheme makes linear operation at the primary user signal as well as the pilot signal of the target band, retain the pilot component at the spectrum holes and convert the recovery of non-sparse primary user signal to the recovery of the sparse pilot signal. The scheme resolves the failure of compressed sensing technology in cognitive radio wideband spectrum sensing when the primary user signal is non-sparse. Simulations indicate that the proposed pilot testing scheme can accurately detect the position of the spectrum hole on non-sparse spectrum.
A multiple description coding transmission strategy proposed to solve the problem of low wireless network connectivity but high transmission quality requirements in wireless multimedia sensor networks (WMSN). This strategy makes multiple descriptions coding on the source node, selects initial broadcasting nodes, and then builds multiple transmission paths. The strategy is applied in image transmission application where the performance is evaluated in terms of data arrival ratio, and comparied between transmitted image, the original one, and peak signal to noise ratio. Experiment shows that the strategy improves wireless network connectivity and assures transmission reliability of WMSN.
In heterogeneous networks where the Macro cell is overlaid by Pico cells,severe downlink interference is suffered by Pico users from the Macro base station of high transmission power. A beam nulling-user selection algorithm is proposed to suppress the interference from the Macro station and therefore enhance the system spectrum efficiency. The interference to Pico users is modeled and analyzed in a mathematical way. Simulation is performed to evaluate the enhancement of the proposed algorithm. It is shown that the performance is significantly improved in case of the high signal-to-noise ratio region or large number of Pico cells.
Aimed at dynamic changing of the available subcarriers in cognitive radio system, a dynamic subcarrier allocation method based on access probability is proposed. The method controls access probability of cognitive users on the subcarriers to maximize the system throughput and guarantee the quality of service requirement of the users. To simplify the solution process, a heuristic algorithm is proposed based on the access probability transfer. The algorithm transfers the subcarrier access probability among the users until the rate requirements of all users being satisfied. Simulation shows that the heuristic algorithm can achieve the global optimal solution in Rayleigh channel environment. Also it is shown that the throughput performance of the proposed method is better than current dynamic allocation and carrier sense multiple access method.
Two blind signature schemes based on the discrete logarithm problem over inner automorphism group are proposed. The schemes satisfy blinding and one-more unforgery. These constructions can be viewed as non-communtative variants of blind signatures. Parameters and performance related issues are also addressed in detail.By comparison, the blind signatures are efficient in the running time and storage space.
For lack of necessary security mechanism, the border gateway protocol (BGP) faces serious security threats. In proposed BGP security mechanisms, the management of complicated certificatation and excessive storage overhead severely block security solutions from being implemented and deployed in real world. Based on modification of autonomous system alliance’s structure, the proxy re-signature is introduced into security enhanced BGP (SE-BGP) to improve translator trust model. An improved BGP security mechanism named improved SE-BGP is also designed for solving the problem of cross-certification of key nodes between autonomous system alliances. Security analysis and performance evaluation demonstrate that this mechanism continues to reduce the route resource expenses to have better scalability and good security capability. The number of used certificates is about 31% of the SE-BGP.
A new traffic transferring based multi-cell load balancing (MLB) scheme is proposed. By multi-cell cooperation, MLB carries out one or two hops traffic transfer among the busy cell and its neighbors and achieves load balancing effectively. A multi-dimensional Markov chain based model is also built for MLB to analyze the system call blocking probability. Simulation verifies that MLB decreases hot cell and system call blocking probabilities greatly comparing to available schemes.
An adaptive service request load-aware cooperative sleeping energy saving algorithm for heterogeneous wireless networks is proposed, and it minimizes system overall power consumption under the constraint on the quality of service requirement in the heterogeneous networks system. The proposed algorithm constructs the economic theory revenue based power-saving model under wireless networks self-optimization framework. Monte Carlo simulation demonstrates that the algorithm can adaptively schedule the working mode of heterogeneous wireless networks and decrease the system total power consumption,and satisfies the blocking rate constraint of user service request.
An adaptive resource optimization algorithm for multiuser orthogonal frequency division multiple access (OFDMA) system is proposed. The algorithm consists of two steps: first, subcarriers are assigned by transiently chaotic neural network (TCNN) when the power is assumpted to be divided equally to every subcarrier. Second, the averagely assigned power is reallocated in a linear water-filling fashion. Simulations show that TCNN has better convergence speed and optimization rate than Hopfield neural network and chaotic neural network. Compared with existing algorithms, the proposed algorithm further achieves higher throughput by exploiting the multiuser diversity gain.
An optimal resource allocation scheme for an orthogonal frequency-division multiple (OFDM) two-way relay network with physical network coding is constructed. Considering the total transmit power and user rate constraints, a sub-optimal resource allocation scheme is proposed to maximize the achievable sum rate of the terminals with low complexity. Using convex optimization techniques, a closed-form solution for power allocation among relay node and two terminals is derived. Thereafter, the variable dimensionality of the objective function is reduced from 3 to 1 that can dramatically decrease the complexity of the solution. Simulation shows that the proposed approach achieves the optimal sum rate and is better than that of the traditional resource assignment method but with less computational load in various scenarios.
To design a post-quantum secure broadcast encryption scheme, by using the bonsai trees model, an identity-based broadcast encryption scheme over lattice is proposed. The proposed scheme can avoid the decryption failure problem in the existing scheme. The new scheme is provable securied against chosen-plaintext attack where the security is reduced to the hardness of learning with errors in the standard model. Moreover the proposed scheme supports the dynamic expansion of client and deals with the setting of stateless receivers.
A quality of experience (QoE) oriented resource allocation scheme for heterogeneous networks (HetNets), termed as MaxQoE, is presented. It can be used in various scenarios with multiple kinds of services in order to provide suitable services quality and allocate appropriate wireless resources for users with different QoE demands. Simulations demonstrate that the proposed scheme can effectively enhance the user QoE in HetNet compared with the maximum throughput scheme and equal allocation scheme..
Based on analysis of joints that interferes with crack detection on cement pavement, a new algorithm is proposed to remove joints and obtain crack identification results by horizontal projection. Firstly, high frequency information of image is obtained by wavelet transform that carries out a binary image containing cracks, joints and noise. Secondly, the joints should be identified and located according to peak features of horizontal projection. Finally, the cracks can be detected accurately after removal of joints. Test illustrats that joints of cement pavement can be removed effectively with this algorithm. Simulation shows that the algorithm can remove the joints effectively, and therefore, the automatic inspecting of cement pavement cracks can be implemented.