Search

Aiming at the problem that the offline screening method of traditional slurry fineness detection is inefficient and cannot be fed back to the upstream grinding system in time, in order to develop the fineness automatic detection technology, an improved surface fitting algorithm called MLTS-LS algorithm (the Moving Least Trimmed Square - Least Square algorithm) is proposed, which performs surface fitting on the slurry fineness data to achieve the purpose of rapid detection of slurry fineness. Firstly, the three-dimensional discrete data of slurry fineness was collected by fineness detection test. Secondly, the surface fitting evaluation indexes of "nearest", "linear", "cubic", "V4" and the traditional least squares method are calculated and analyzed, and an improved interpolation algorithm proposed. Finally, the "MLTS-LS" algorithm is applied to the fitting of three-dimensional discrete data of slurry fineness. The results show that the sum variance and mean square deviation values of the "MLTS-LS" algorithm are significantly smaller than those of other algorithms, and the determination coefficient value and the correction coefficient value are close to 1, indicating that the "MLTS-LS" algorithm has a good fitting effect on the three-dimensional discrete data of slurry fineness. [ABSTRACT FROM AUTHOR]

Accurate automatic segmentation methods for liver and liver tumors are important in helping physicians to diagnose, treat, and observe liver cancer in the postoperative period. Due to the intrinsic locality of convolution, existing convolution-based methods are difficult to establish long-range dependencies. Transformer's cascading attention mechanism can establish global information association but will destroy local details. Based on this, a feature modeling method that fuses convolution and Transformer is proposed. The method interactively fuses local and global representations by mixed embedding to maximize the global dependencies at different resolutions. Meanwhile, the contextual information from different encoding stages is captured by multi-level feature fusion module at the skip connection to obtain richer semantic information. Finally, in order to cope with the variation of liver tumors in size and shape, a deformable multi-scale module is used to extract multi-scale features of tumors. The experiments mainly use Dice similarity coefficient (DSC) as evaluation metrics. The DSCs of liver and tumor on the LiTS17 dataset are 0.920 and 0.748, respectively, and the results show that the proposed network has more accurate liver tumor segmentation results compared to the baseline. [ABSTRACT FROM AUTHOR]

In the algorithm of using convolutional neural network to identify verification codes, the traditional algorithm randomly initializes the convolutional kernel parameters, which leads to slow convergence speed, long training time and low recognition accuracy. To solve this problem, this study proposed a verification code recognition method based on principal component analysis and convolutional neural network. The algorithm first used the drop-fall algorithm to segment the preprocessed verification code image, and then lused PCA learning for the segmented image, and took the learned feature vector as the initialization parameter of the convolution core of the corresponding convolution layer. The experimental results show that the improved method has better performance in convergence speed and recognition accuracy. [ABSTRACT FROM AUTHOR]

Anaplastic lymphoma kinase (ALK) is a receptor tyrosine kinase and its rearrangements occur in non- small cell lung cancer (NSCLC), resulting in signal dysregulation in kinase domain. As a new generation of potent ALK tyrosine kinase inhibitors (TKIs), Brigatinib was approved in China in March 2022 as a treatment for locally advanced or metastatic NSCLC patients with ALK rearrangement positive. Brigatinib significantly improved the survival, cranial efficacy and quality of life compared to Crizotinib in clinical trials. Brigatinib is generally well tolerated. Brigatinib has been one of the preferred treatments and an addition of options in ALK-rearranged NSCLC. Pulmonary toxicity is one of the adverse effects observed during the treatment of TKIs and deserves the intense attention of clinicians, despite of its low incidence rate. Pulmonary toxicity reported during the treatment of Brigatinib has shown distinct clinical presentations, such as early-onset (median time to onset, 2 days) and rapid tolerance and reversibility of symptoms. In view of this, the concept of early-onset pulmonary events (EOPEs) was proposed and established during the submission for regulatory review and approval. We focused on clinical characteristics, potential mechanism of etiology, and management strategies of EOPEs to provide clinicians evidence for better clinical decision support. [ABSTRACT FROM AUTHOR]

With the rapid development of artificial intelligence (AI), machine learning (ML) algorithms has been widely used in stroke diagnosis, treatment and prognosis, and has shown great application potential. This paper summarizes the application status of ML algorithms in stroke research from the aspects of ML algorithms research, ML algorithms in stroke diagnosis and classification diagnosis, outcome prediction, risk factor identification and so on, and looks into the future development direction. [ABSTRACT FROM AUTHOR]

To solve the problem of difficult focus, serious geometry distortion and large cross sidelobe in high-squint mode synthetic aperture radar (SAR) ground moving target imaging (GMTIm), a high-squint SAR GMTIm algorithm based on the ωk algorithm with spectrum rotation is proposed. Firstly, a moving target echo signal model with accurate slant-range is deduced in detail to obtain preliminary imaging result and extract the region of interest (ROD moving target data. Then, the phase compensation function is constructed and the minimum entropy algorithm is utilized to finish moving target parameter estimation. According to the estimated parameter, the phase of ROI data is compensated to get the focused moving target image. Furthermore, spectrum rotation eliminates geometry distortion, and the sparsity enhancement method decreases the moving target sidelobe. By doing so, focused, no geometry distortion and low sidelobe moving target image is obtained. Simulation results validate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Aiming at the singularity of real-time model transition probability correction using model likelihood function comparison with traditional interacting multiple model (IMM) algorithm, an improved adaptive IMM algorithm is proposed based on the proposed correction function. Firstly, combining the white noise model with the extended Kalman filter (EKF) algorithm, the non maneuvering model EKFI and maneuvering model EKF2 are designed as the model set of IMM algorithm. Secondly, the prediction model adopts the nonlinear relative orbit dynamic equation suitable for the elliptical reference orbit to improve the prediction accuracy of the model. Finally, the effect of rate measurement information on reducing the peak error of maneuvering target tracking is analyzed. The simulation results show that the tracking accuracy of the improved model transition probability adaptive imm-ekf algorithm is significantly improved and better than the existing methods. After introducing the rate measurement information, the maximum peak error and estimation accuracy are improved. [ABSTRACT FROM AUTHOR]

Aiming at the problems of low accuracy, slow speed and low defect identification of industrial site fabric defect detection, a improved fabric defect detection algorithm based on ResNet50 network is proposed. Firstly, the data set was preprocessed, and the model training set was enhanced by cutting the data samples, including the fault-free and 8 common defect categories. Then, the structure of Resnet50 network was improved to extract the weight parameters pre-trained on the large data set ImageNet for transfer learning. Finally, the defect detection and recognition model obtained by hyper parameter training was adjusted repeatedly. The results of several groups of comparative experiments show that the recognition accuracy of the improved model for normal fabric and 8 common defects reaches 96. 32%, which is 4.2% higher than that of the standard model, and the speed is doubled. The test in different fabric defect data sets has the best comprehensive performance, strong generalization ability and good robustness, which can meet the requirements of fabric defect detection in industrial production sites. [ABSTRACT FROM AUTHOR]

Aiming at the problem of slow speed on convergence and poor separation perfomance for the method of traditional independent component analysis (ICA). The strategy on chaotic maps strategy are combined with an adaptive explosion radius, and a kind of chaotic adaptive fireworks algorithm (CAFWA) based blind source separation (BSS) method is proposed, and applied into the sorting problem of radar emitter mixed signal. The chaotic maps strategy can distribute the initial value more evenly in the solution space, and the explosion radius can be changed adaptively according to the advantages and disadvantages of fitness, which ensures the precision of local search for the proposed algorithm and satisfies the diversity of universal search. The experimental results show that the proposed algorithm can sort observation signals well in both noise-free and noise conditions, and has faster constriction speed and better sorting function than the traditional algorithm. [ABSTRACT FROM AUTHOR]

Aiming at the combat situation when the handover area of hypersonic target interceptor is high in the near space, referencing to the theory of orbit interception, based on the high throw and reentry interceptor trajectory and take interceptor as two body model, a long-range interceptor guidance algorithm is designed. Firstly, the complex motion model is simplified, and according to the force analysis, the model is transformed into a two body orbit model. Then, by using the track angle iteration method, the Lambert problem is solved, and the orbit change required speed is obtained. It is proposed to adjust the speed to the required speed of the orbit change to complete the intercepting task in the orbit model. Finally, because of the limited thrust, the maneuverability of the interceptor in transition period is analyzed, and the method of velocity gain guidance is proposed to solve the problem. The simulation results show that it is feasible to use the two body orbit model to represent the motion model of the interceptor under the condition of weak aerodynamic force in the high altitude domain. The speed gain guidance algorithm can not only solve the transition period maneuver problem effectively, but also has good convergence in predicting the change of interceptor points, so as to complete the medium guidance interception task in the near space. [ABSTRACT FROM AUTHOR]

A new planar sparse geometry, referred to as the sparse H array was proposed by re-optimizing the sensors of open box array (OBA) based on difference co-array concept. The new structures has the same number of identifiable sources as the OBAs. It can decrease the mutual coupling effect between sensors by effectively reducing sensor pairs with closed space. Simulations of the direction-of-arrival estimation verify the better anti-coupling capability of the proposed PSAs than other geometries with the 2-D unitary ESPRIT algorithm used. [ABSTRACT FROM AUTHOR]

In high dynamic and long-term integration condition, the code Doppler frequency will lead to code phase migration in PMF-FFT algorithm, which is not conducive to the DSSS signal acquisition. After in-depth study on this issue, the quantitative relationship among the code phase migration, integration losses and parameters estimated bias were derived and simulated. Based on PMF-FFT algorithm, a compensation method for code Doppler frequency was proposed, which used Keystone transform to correct code phase migration, so as to compensate the code Doppler frequency. Simulation results show that, on the condition of 37 dB/Hz CNR, 10 ms coherent integration interval and a fixed false alarm possibility of 10-6, the new algorithm can effectively eliminate the code Doppler frequency and has a single detection possibility of 98%. [ABSTRACT FROM AUTHOR]

To meet the requirement of various resolution displays on mobile terminals, an improved set portion in hierarchical trees (SPIHT) image coding algorithm is proposed, which integrates the seam carving (SC) technique. In the presented method, to generate spatial scalable bit stream, a new structure of spatial orientation tree (SOT) based on root nodes is constructed and the SC is used as a guide for SPIHT coding procedure. Mobile receivers can be capable of finishing the compression decoding and image resizing simultaneously only requiring the corresponding bit stream by using the proposed algorithm. Experimental results show that when the target resolution of receiver is the same as the original resolution, the average rate-distortion ( R-D) performance of the proposed method is much close to that of the traditional SPIHT coding algorithm. While the target resolutions are lower than the original resolution, the proposed algorithm is superior to the traditional SPIHT coding approach whatever for coding bit rates or subjective quality of reconstructed images. [ABSTRACT FROM AUTHOR]

Agricultural mobile robot and sightseeing agriculture is a direction of agricultural development in recent years. Agricultural mobile robot, a kind of efficient transportation equipment and means of transport, was of great significance in the orchard sightseeing agriculture. Road detection is the key technology and an important prerequisite for mobile agricultural robot to achieve autonomous navigation. In practical applications, the complexity of the orchard environment, e.g., the impact of illumination changes, shadows and occlusion, has resulted in poor robustness of vision detection algorithm. Therefore, the orchard road detection algorithm is required to be improved. So a method fusing edge detection and improved random sample consensus for winding orchard path detection was proposed. The proposed algorithm was consisted of orchard road edge detection algorithm (REE) and improved RANSAC algorithm (IRANSAC). Because the orchard road image contained a lot of noise, such as shadows and occlusion, the REE was aimed at extracting road edge as well as removing noise according to the color distribution and geometry characteristics of the orchard road. First, using the finite difference operator to extract image edge may contain noises. Then a basic assumption that road edges had striking gray contrast among their neighborhood was proposed, so we used the constraint of contrast of gray values to removed noises. However, some noises satisfied the constraint condition, hence another assumption that a curved road could be seen as straight road in a certain scale was proposed, therefore, the image was divided into n regions, if n was large enough, a linear curve could approximate to curve in sub-image. On this basis, an improved hough line detection algorithm was executed to remove noises which were not lying on the lines. The REE could dramatically remove noises and keep the road edge points. However the REE could not remove all the noises, so the linear segments in the image could not represent curve. The spline curve model was proposed to describe the line or curve road, so the remaining problem was how to find the true spline among the edge points. IRANSAC was aimed at fitting the spline curve. The IRANSAC combining the advantages of linear least square method and RANSAC could correctly estimate model parameters of the spline curve, and achieve detecting orchard road. In order to test the proposed algorithm, we collected 240 Orchard Road images as test objects, including straight roads, curved roads, roads disturbed by illumination change and blocked roads in the South China Agricultural University. The result showed that: under the influence of illumination change and occlusion, the REE algorithm can effectively extract the edge of orchard image, and reduce 96.5% residual noises effectively, with the average computation time of 0.1658 s; The IRANSAC can correctly fit the road edge, and the correct fitting rate of the four roads are 93.3%, 86.7%, 85.0% and 91.7% higher than RANSAC respectively, with the average correct rate of 89.1% and the average detection time of 0.1834 s; Sometimes the IRANSAC failed to fit the right spline curve because the complex environment may cause road edge points missing, or the REE algorithm failed to get sufficient edge points. In brief, the proposed algorithm can satisfy the robustness of navigation system and real-time requirements, and ensure the effectiveness of the visual navigation system to achieve orchard road detection. In order to further improve the algorithm robustness under the clutter background, the key point is to improve robustness of REE. [ABSTRACT FROM AUTHOR]

A new approach of monocular visual odometry based on the Kinect sensor and an improved SURF feature extraction algorithm is presented in this paper. The color and depth image are obtained by the Kinect sensor. The feature points are extracted by the improved SURF algorithm based on feature points information to improve the matching accuracy and robustness. The feature points are mapped into depth image, and least median squares method is well used to estimate the path information of the robot. The experiments show that this method can reach a higher accuracy than SURF and has a strong robustness in dynamic environment. This method is a simple and effective new method. [ABSTRACT FROM AUTHOR]

Due to the failure-prone nature of communications, this paper proposed a fault-tolerant and privacy-preserving data aggregation algorithm, which employed an elliptic curve cryptography-based additively homomorphic encryption algorithm to offer end-to-end privacy of sensing data, and then savd energy and bandwidth. Meanwhile, construction track graph topology, each node had more than one parent node, so that when a link error in the primary edge, the backup parent could recover the correct aggregate values. The theoretical analysis and simulation results illustrate this scheme has good privacy of data, low communication overhead and high precision aggregate values. [ABSTRACT FROM AUTHOR]

To find the optimal scheduling for the production job in the production process, using the production quality, efficiency and cost threshold value as constraints, this paper built a multi-objective relative degree of nearness optimization model for quality-efficiency-cost control based on set pair analysis (SPA). It solved the optimization model by using non-dominated sorting generic algorithm (NSGA-II) and acquired the Pareto-optimal solutions under the relative certainty conditions. Each job selected reasonable solution according to the optimal alternative. Finally, the experiment verifies the multi-objective optimization problems that illustrates the accuracy, effectiveness and applicability of the proposed method. [ABSTRACT FROM AUTHOR]

The computational complexity for the Petri nets control problem is very high due to the exponential growth of the state space for the uncontrollable subnets. To overcome this complexity problem, an algorithm is proposed to equivalently transform the original control problem into a much simpler one. This algorithm includes two parts. One is to compress an area of the uncontrollable subnet into a single place. The other is to equivalently transform the original linear constraint into a new one on the compressed net. Since an uncontrollable subnet can be effectively compressed by iterating this algorithm, and the state space from any marking can be exponentially reduced as the net size decreases, this algorithm can effectively reduce the computational complexity of the control problem. Especially when the uncontrollable subnet is a state machine, this method can be used to design an efficient and optimal supervisor. A material transportation system illustrates this method. [ABSTRACT FROM AUTHOR]

To establish the universal model of repairable spare parts inventory distribution, considering relative important repair parameters such as "repair in place rate", "duty cycle" and the "retest ok rate", the initial inventory distribution model for multi-indenture spare parts under the multi-echelon maintenance supply system is established based on the METRIC theory. The equipment availability, spare parts fill rate and support delay are used as the optimization target while the cost is taken as the constraint to optimize the inventory distribution project. The optimization result is got through a given example, and the support factors are analyzed. The result is tested by VMETRIC simulation software, which shows that the two experiment results are nearly identical, so the model's correctness and the algorithm's rationality are proved. [ABSTRACT FROM AUTHOR]

Centipede-like robot, which has high degree of freedom and similar body segments, proposes new requirements on real-time control for diverse gaits. To meet above requests, we have studied the movement characteristics of multi-legged creatures and found the features of gait periodic relay. Then the instruction-relay control scheme, the three-bus control system and its software algorithms were designed. In our experiments, with the designed control system, different gaits of robot could be switched flexibly and the phase difference between body sections could be changed by altering relay time. As a whole, this control system could accomplish the requested control task and reasonably simplify the gait control algorithms and procedures. [ABSTRACT FROM AUTHOR]

Both insurer's default risk and insured's surrender events are important factors in pricing unitized participating life insurance. In order to extend the results presented in our earlier research, we suppose that both surrender and default could happen at the end of each insurance year. In this paper, three models are constructed for pricing surrender option with insolvency. An iterative algorithm based on Monte Carlo simulation method is then implemented in order to compute them. The findings are as follows: price of unitized participating life insurance could be badly underestimated without consideration of surrender option, price of which is highly sensitive to some exogenous variables such as volatility. It was also implied that growth in default risk tends to result in growth in value of surrender option. [ABSTRACT FROM AUTHOR]

To conquer the disadvantages of low efficiency and the poor practicability of the traditional attack tree, a universal serial attack model and its constructing method were proposed by introducing an order of the atomic attacks. The weights of each tree leave were defined and the effectiveness of the elementary attacks could be quantified. Further, the Dijkstra algorithm was used to calculate the optimal route of the serial attack model, which would offer an optimal attack strategy. Finally, a complex attack tree with twenty leaves was built and analyzed by using the above method. The results indicate that, in comparison with the traditional parallel model, the proposed serial attack tree model with weight concept has the advantages of bigger outcome, lower algorithm complexity, higher efficiency and more perfect effectiveness. [ABSTRACT FROM AUTHOR]

This article presents a study in order to solve the problem that the phase estimation cannot complete the application for weak GPS signal. Phase doubled processing and decision directed feedback were both used to eliminate the influence of the reverse of navigation data. A new maximum likelihood (ML) phase estimator based on incoherent integration with decision directed feedback was proposed. The expressions of SNR were derived in theory. RMSE was acquired by numerical integration of the conditional probability density function (CPDF). The new method could reduce RMSE of ML phase estimator and improve the capability of phase estimation for weak GPS signal. The effectiveness of theory analysis and the proposed algorithm were proved by simulation. The practicability of the algorithm is indicated by the result of the implementation based on field-programmable gate array. [ABSTRACT FROM AUTHOR]

Setup (changeover) time is needed by the processes with cleaning or changing fixtures operations, which is strongly dependent on the immediately succeeding operation. The scheduling problems in this environment are sequence-dependent. This paper focuses on the single machine scheduling problem with sequence dependent setup times to minimize the total weighted tardiness, which is NP-hard. An iterated filter-and-fan algorithm (IFF) is proposed to solve it, which searches the solution space as a manner of a branch tree. In this algorithm, when the branching moves cannot improve the root solution, a new root solution with inherited features is generated, from which the search continues to work. According to the solution characteristic of the problem, a branching move strategy with sequence segment recomposing and referenced local search is proposed to generate branch nodes. The algorithm is tested by 120 instances from literature. The results show the solutions obtained by the IFF algorithm are better or not worse than those obtained by DE and DPSO algorithm for most instances. Besides, the best solutions for 42 instances are improved by the IFF algorithm. [ABSTRACT FROM AUTHOR]

A GMRES algorithm for hovering rotor simulation is presented. Two calculation methods of matrix-vector product in this algorithm are discussed. The Euler equations with absolute variables on rotating coordinates are directly solved by this way. The flow solver utilizes unstructured mixed grid and cell-vertex finite-volume scheme based on the Upwind Roe's approximate Riemann solver. Using this method a transonic hovering flow fields of Caradonna-Tung rotor are simulated. The computed results show good agreement with experimental results. The efficiency of the GMRES algorithm and LU-SGS algorithm is compared. The comparing results show that the GMRES algorithm can remarkably speed up the convergence rate of the rotor flow. [ABSTRACT FROM AUTHOR]

For the better use of metamaterials, the surface waves of metamaterials are studied in half-space model with dipole radiation source. Using the Sommerfeld integrals of TM wave and TE wave, the branch-cut's effect on the computation of the integrals is investigated based on saddle point algorithm. The distribution and validity of pole points for the integrands is analyzed according to the properties of complex integral. For TM wave, there are lateral wave and Zenneck surface wave in the case of taking left-handed material as substrate; for TE wave, no Zenneck surface wave can exist in any case of metamaterials. [ABSTRACT FROM AUTHOR]

Under the guidance of the simulation of antigen-antibody recognition in biological mechanisms of the immune system, based on the artificial immune algorithm, a prediction model for slope stability was introduced. The slope stability sample set was defined as antigen set, and the influence factors of slope stability were defined as the antibody. Through reiteration of genetic manipulation on the antigen gene segment, the antibody set that can perform slope stability well was developed. The affinity between prediction sample set and antibody set were calculated, and the KNN algorithm was used to predict the sample stability. The self-adaptive artificial immune algorithm was also employed to improve the model effectiveness and reliability. The case study confirms that self-adaptive artificial immune algorithm has a more accurate prediction than basic artificial immune algorithm, proving that the self-adaptive approach is effective. The new method can avoid from building complex non-linear function between influence factors and stability, effectively reduce the modeling complexity and has better adaptability. [ABSTRACT FROM AUTHOR]

For the accuracy of indoor location system based on the strength of received signal being restricted by the multipath effect and non line of sight propagation, a location algorithm using radio frequency identification was proposed which was suitable for 915 MHZ UHF signal. Dual frequency subcarriers were introduced to overcome the ambiguity of whole cycles and reduce the demands of sample rate. On the basis of spectrum analysis of All-Phase FFT, a mechanism was put forward to estimate the phase difference so as to achieve the ranging information with under-sampling, and then the position of the tracking tags could be obtained by the least square method. Simulation results show that, the proposed algorithm possesses a higher accuracy and stability. [ABSTRACT FROM AUTHOR]

To optimizing coalbed methane (CBM) favorable targets in western Guizhou Province and eastern Yunnan Province, six typical areas were focused on to estimate which should be explored preferentially. Firstly, nine parameters in three categories, e. g. the coal reservoir, the resource conditions and the exploitation infrastructures were concentrated to form an evaluation matrix. Subsequently, the matrix was weighted by a vector that combined the grey correlation analysis and the analytic hierarchy process (AHP) and was calculated in a fuzzy multi-objective and multi-level optimization model. Finally, the results of the calculation were drawn on a thematic map. It indicate that the optimal target is the Zhina Coalmine, and the Panxian Coalmine is suboptimal as well as the Laochang, Liuzhi and Shuieheng Coalmines. By contrast, the Enhong Coalmine is unfavorable. [ABSTRACT FROM AUTHOR]

In order to recognize the criminal suspects in the crowd relevant to the case by computer and improve the efficiency of solving the case, this paper presents a new method of suspected culprit recognition based on Probit model. Separation algorithm, correlation algorithm based on cluster and the parameters named significance level in Probit model are adopted to find the important attributes of the culprit, and then, by training the data just including the important attributes, the crime risk evaluation model could be achieved. The experimental results show that average accuracy of model-recognizing-culprit is 90.5% and the average recall is 92.7%. [ABSTRACT FROM AUTHOR]

The traditional particle filter tracking algorithm usually leads to tracking error or failure, when the target is interfered by the similar background or blocked by the other object. To eliminate these effects, an improved method was proposed. To overcome the background interference, the logarithm likelihood function was used to fuse the background information into the target model, and the target was divided into multiple sub-regions to fuse the spatial information into the target model, which increase the model reliability. Considering that the target might be under occlusion, the historical trajectory information was stored over a little of time and the least square method was used to predict the target location in the next frame. Experimental results show that the improved tracking method is more robust to background interference and has better tracking accuracy in occlusion cases than the traditional particle filter. [ABSTRACT FROM AUTHOR]

In this paper, we proposed a novel image hash algorithm based on Zernike moments. We employ Zernike moments to represent the image content for the reason that it is the projection of the image under a group of orthogonal basis. Thus, the discrimination of proposed algorithm is improved. Meanwhile, benefited from the invariance property of Zernike moment under affine transform, our algorithm is robust to rotation attack. In the proposed method, the hash string is generated from the Zernike moments of the normalized input image. Secret key is also introduced to ensure the security of the hash method. In the experiments, the method is tested by 100 sample images, The false accept rate under rotation attack is less than 4%. The error recognizing rate between different images is about 10-9. Experimental results prove that our algorithm has great robustness and discrimination. [ABSTRACT FROM AUTHOR]

Aimed at the problem that the mean shift algorithm sometimes can not compute large volumes of image data, or the data-computing speed may be very slow, by analyzing the principle of the mean-shift image segmentation, this paper presents an image-segmentation idea by processing the data in parallel computing environment, i.e. Data Partition Method with Buffer (DPMB), then we can compute different parts of data separately. By this method, we can avoid the limit of data amounts and memory errors problem of computer, which improves the efficiency of data segmentation. [ABSTRACT FROM AUTHOR]

To improve the transmission efficiency and obtain the information of Channel in OFDM system, a novel nonlinear channel prediction algorithm is proposed. Firstly, Support Vector Machines were associated with nonlinear dynamics using the theory of reconstructed embedding phase space to determine the dimension of training data set, which provides a novel approach to investigate the Support Vector Machines. Then the prediction of pilot tones in time field was realized by Recurrent Least Squares Support Vector Machines (RLS -SVM) in high dimensional space, and the interpolation operation was carried out in frequency field to estimate parameters of other tones. Simulation results show that the number of pilot tones is decreased at the same transmission rate and the transmission efficiency is improved under improvement of bit error rate. In addition, this method provides important Channel State Information (CSI) for realizing adaptive techniques. [ABSTRACT FROM AUTHOR]

The shortcomings of adaptive control arise especially in the case of the systems with significant uncertainty on plant parameters, such as large overshoots in the transient stages, and the traditional adaptive controller based on certainty equivalence (CE) assumption and cautious controller can not provide good control quality for these systems with unknown and time-varying parameters. To overcome the problem of large moment of inertia changes of the multi-joint mechanical arm in motion, a model reference adaptive dual control algorithm is presented and used in the pose control of mechanical arm. The simulation and experimental results show that the performance of dual control is superior to that of the traditional adaptive control based on the CE assumption and the cautious control. [ABSTRACT FROM AUTHOR]

A personality music recommendation algorithm model based on Mel-frequency cepstrum coefficients(MFCC) and Gaussian mixture model (GMM) is provided. This method extracts MFCC from a certain song as feature parameters, and generates a template of the song using the GMM algorithm. It then gains similar songs from the music library by comparing their templates' through similarity. From the experimental result, the correct rate of the song recommendation is 90%. [ABSTRACT FROM AUTHOR]

Investigates a signal processor of electromic warefare(EW) digital receiver based on FFT/IFFT algorithm, and realized completely by FPGA constructed annularly. Four FPGAs implement respectively the high-speed data transmission interface, FFT/IFFT operation and signal detection in the time/frequency domain. The FPGAs run with distributed, multi-bus, parallel and pipeline mode. This processor can detect the parameters of carrier frequency and pulse description word for 4 simultaneous-arrival signals at most, and they need only about 20 ms to detect 4 signals while 256 K(1 K= 1 024) point FFT and 32 K point IFFT are adopted. The data receiving, operating and signal detection in time/frequency domain are realized on a board. [ABSTRACT FROM AUTHOR]

Due to the difficulty of estimating probability function and the high space complexity of the existing probabilistic neural network (PNN), an improved PNN model is presented by introducing the mechanism of back-propagation (BP). The improved model inherits the principle and structure of PNN, and meanwhile applies the BP algorithm of multilayer perceptron (MLP) to train probability function and parameters. The above two aspects help overcome the PNN's shortcomings. Three numerical experiments have been designed to verify the improved model, and their results indicate that the new model has strong capacity to identify the importance of input indicators and own high accuracy of classification. In conclusion, the BP-PNN model is a new pattern classification method with widespread applicability and can support management decision. [ABSTRACT FROM AUTHOR]

The ultimate goal of a steganography algorithm is to carry as many bits as possible without introducing any detectable artifact into the cover object. However, embedding quality and payload have an antagonistic relationship. Then, people often make a tradeoff between them depending on their different desires. In this paper, a new steganography algorithm for grayscale images with large payload and acceptable visual effect is proposed. We use a group of functions to guide cover pixel pairs' modification and imply two secret digits base-9 numeral system one time. From the experimental results, the embedding capacity of the proposed algorithm achieves 3.1699 bit per pixel, which is much larger than that of the published works, while the visual quality of stego images is acceptable too. [ABSTRACT FROM AUTHOR]