Email alert
2015 Vol. 37, No. 12
Display Method:
2015, 37(12): 2795-2801.
doi: 10.11999/JEIT150422
Abstract:
The type and scale of data has been promoted with a hitherto unknown speed by the emerging services including cloud computing, health care, street view services recommendation system and so on. However, the surge in the volume of data may lead to many common problems, such as the representability, reliability and handlability of data. Therefore, how to effectively handle the relationship between the data and the analysis to improve the efficiency of classification of the data and establish the data clustering analysis model has become an academic and business problem, which needs to be solved urgently. A hierarchical clustering method based on semantic feature is proposed. Firstly, the data should be trained according to the semantic features of data, and then is used the training result to process hierarchical clustering in each subset; finally, the density center point is produced. This method can improve the efficiency and accuracy of data clustering. This algorithm is of low complexity about sampling, high accuracy of data analysis and good judgment. Furthermore, the algorithm is easy to realize.
The type and scale of data has been promoted with a hitherto unknown speed by the emerging services including cloud computing, health care, street view services recommendation system and so on. However, the surge in the volume of data may lead to many common problems, such as the representability, reliability and handlability of data. Therefore, how to effectively handle the relationship between the data and the analysis to improve the efficiency of classification of the data and establish the data clustering analysis model has become an academic and business problem, which needs to be solved urgently. A hierarchical clustering method based on semantic feature is proposed. Firstly, the data should be trained according to the semantic features of data, and then is used the training result to process hierarchical clustering in each subset; finally, the density center point is produced. This method can improve the efficiency and accuracy of data clustering. This algorithm is of low complexity about sampling, high accuracy of data analysis and good judgment. Furthermore, the algorithm is easy to realize.
2015, 37(12): 2802-2807.
doi: 10.11999/JEIT150356
Abstract:
In order to solve the selfishness problem that forwarding nodes in the wireless ad hoc network refuse to cooperation due to the limit of energy and storage space, a wireless ad hoc network cooperation enhancement model combined with the game theory is proposed, which is based on the incentive mechanism of virtual currency, analysis the benefit and overhead characteristics of the source nodes and forwarding nodes. In this model, the network cooperation problem is transformed into a game equilibrium problem about the benefit of source node and forwarding nodes in the data forwarding path, promoting the cooperation of communication. Furthermore, in order to avoid the congestion and maximizing network lifetime, the model makes some certain constraint about the energy and storage space for the forwarding nodes.
In order to solve the selfishness problem that forwarding nodes in the wireless ad hoc network refuse to cooperation due to the limit of energy and storage space, a wireless ad hoc network cooperation enhancement model combined with the game theory is proposed, which is based on the incentive mechanism of virtual currency, analysis the benefit and overhead characteristics of the source nodes and forwarding nodes. In this model, the network cooperation problem is transformed into a game equilibrium problem about the benefit of source node and forwarding nodes in the data forwarding path, promoting the cooperation of communication. Furthermore, in order to avoid the congestion and maximizing network lifetime, the model makes some certain constraint about the energy and storage space for the forwarding nodes.
2015, 37(12): 2808-2814.
doi: 10.11999/JEIT150208
Abstract:
To address the contradiction between data aggregation and data security in Wireless Sensor Networks (WSNs), a recoverable privacy-preserving integrity-assured data aggregation scheme is proposed based on the technologies of privacy homomorphism and aggregate message authentication code. The proposed scheme enables the Base Station (BS) to recover all the original sensing data from the final aggregated results, which makes it possible to verify the integrity of each sensing data and aggregated data, and perform any further operations on them on demand. The security analysis shows that the proposal not only provides the data privacy and data integrity, but also resists against unauthorized aggregation attack and aggregator capture attack; besides, it is able to detect and locate the malicious nodes which injects false data to the network in a certain range. The performance analysis shows that the proposed scheme has remarkable advantages over existing schemes in terms of computation and communication overhead. In order to evaluate the performance and feasibility of the proposal, the prototype implementation is presented based on the TinyOS platform. The experiment results demonstrate the proposed scheme is feasible and efficient for resource-constrained WSNs.
To address the contradiction between data aggregation and data security in Wireless Sensor Networks (WSNs), a recoverable privacy-preserving integrity-assured data aggregation scheme is proposed based on the technologies of privacy homomorphism and aggregate message authentication code. The proposed scheme enables the Base Station (BS) to recover all the original sensing data from the final aggregated results, which makes it possible to verify the integrity of each sensing data and aggregated data, and perform any further operations on them on demand. The security analysis shows that the proposal not only provides the data privacy and data integrity, but also resists against unauthorized aggregation attack and aggregator capture attack; besides, it is able to detect and locate the malicious nodes which injects false data to the network in a certain range. The performance analysis shows that the proposed scheme has remarkable advantages over existing schemes in terms of computation and communication overhead. In order to evaluate the performance and feasibility of the proposal, the prototype implementation is presented based on the TinyOS platform. The experiment results demonstrate the proposed scheme is feasible and efficient for resource-constrained WSNs.
2015, 37(12): 2815-2820.
doi: 10.11999/JEIT150191
Abstract:
Matching prediction with high accuracy of entity state can cansignificantly improve the efficiency of content-based search in the Internet of Things and reduce communication overhead while searching. The equal-interval and during the period entity state prediction methods are proposed, which are applied to the estimation of the entity state at the moment of querying. Moreover, the ordered verification approach is designed to verify the entities in sequence based on the degree of compliance with the searching content, for the sake of enhancing the reliability of searching results. Numerical results show that the proposed entity state prediction approachescan achieve high accuracy, which combines with the ordered verification approach to dramatically improve the performance of communication overheadduring the searching process.
Matching prediction with high accuracy of entity state can cansignificantly improve the efficiency of content-based search in the Internet of Things and reduce communication overhead while searching. The equal-interval and during the period entity state prediction methods are proposed, which are applied to the estimation of the entity state at the moment of querying. Moreover, the ordered verification approach is designed to verify the entities in sequence based on the degree of compliance with the searching content, for the sake of enhancing the reliability of searching results. Numerical results show that the proposed entity state prediction approachescan achieve high accuracy, which combines with the ordered verification approach to dramatically improve the performance of communication overheadduring the searching process.
2015, 37(12): 2821-2829.
doi: 10.11999/JEIT150387
Abstract:
As current kernel malware detection method based on data signature exists the problem that its efficiency decreases with the growth of the number of signatures, a signature selection method for kernel malware based on hierarchical cluster is presented. First, since current similarity calculation methods are difficult to be applied to data signature selection, a longest common subset based method and a 2-round Hash computation algorithm are introduced. Second, a longest common subset based hierarchical cluster algorithm is presented, thereby performing similar signature aggregation effectively. Finally, a signature selection algorithm based on inconsistent coefficient is designed to reduce the number of signatures. Experimental results show the effectiveness of the method, and performance evaluations indicate that algorithm runtime is acceptable.
As current kernel malware detection method based on data signature exists the problem that its efficiency decreases with the growth of the number of signatures, a signature selection method for kernel malware based on hierarchical cluster is presented. First, since current similarity calculation methods are difficult to be applied to data signature selection, a longest common subset based method and a 2-round Hash computation algorithm are introduced. Second, a longest common subset based hierarchical cluster algorithm is presented, thereby performing similar signature aggregation effectively. Finally, a signature selection algorithm based on inconsistent coefficient is designed to reduce the number of signatures. Experimental results show the effectiveness of the method, and performance evaluations indicate that algorithm runtime is acceptable.
2015, 37(12): 2830-2837.
doi: 10.11999/JEIT150477
Abstract:
To make the generalized interger transform used for lossless information hiding more generalized and improve the data embedded capacity, a new lossless information hiding method based on expanded generalized integer transform is proposed for image. By introducing the variable parameterm and expanding the generalized integer transform, m(n-1) bit data can be embedded into one image block withn pixels, improving the embedded capacity whenm＞1. Besides, 5 the method uses a new data embedding strategy to improve the quality of the hidded image. Compared with current integer transform algorithms, there is a certain improvement in the embedded capacity and the quality of hidded image with the proposed algorithm. The proposed method can be used to some important and sensitive areas, i.e. military comunication, healthcare, and law-enforcement, after extracted the embedded secret data, the cover image can be recovered losslessly.
To make the generalized interger transform used for lossless information hiding more generalized and improve the data embedded capacity, a new lossless information hiding method based on expanded generalized integer transform is proposed for image. By introducing the variable parameterm and expanding the generalized integer transform, m(n-1) bit data can be embedded into one image block withn pixels, improving the embedded capacity whenm＞1. Besides, 5 the method uses a new data embedding strategy to improve the quality of the hidded image. Compared with current integer transform algorithms, there is a certain improvement in the embedded capacity and the quality of hidded image with the proposed algorithm. The proposed method can be used to some important and sensitive areas, i.e. military comunication, healthcare, and law-enforcement, after extracted the embedded secret data, the cover image can be recovered losslessly.
2015, 37(12): 2838-2844.
doi: 10.11999/JEIT150407
Abstract:
Certificateless aggregate signcryption not only can ensure the confidentiality and authentication of information transmission, but also can reduce the cost of data communication and the verification of ciphertexts. Through analyzing some existing certificateless aggregate signcryption schemes, it is found that their efficiencies are much lower. A provable secure certificateless compact aggregate signcryption scheme is proposed in this paper. In the new scheme, the pairing numbers, not depending on the number of signcryption users, are constant when aggregate ciphertexts are verified. Compared with the existing certificateless aggregate signcryption schemes, the new scheme decreases pairing numbers and raise the efficiency of verification. Moreover, based on the assumption of bilinear Diffie-Hellman and computational Diffie-Hellman, in the random oracle model, it is proved that the new scheme satisfies the properties of confidentiality and unforgeability.
Certificateless aggregate signcryption not only can ensure the confidentiality and authentication of information transmission, but also can reduce the cost of data communication and the verification of ciphertexts. Through analyzing some existing certificateless aggregate signcryption schemes, it is found that their efficiencies are much lower. A provable secure certificateless compact aggregate signcryption scheme is proposed in this paper. In the new scheme, the pairing numbers, not depending on the number of signcryption users, are constant when aggregate ciphertexts are verified. Compared with the existing certificateless aggregate signcryption schemes, the new scheme decreases pairing numbers and raise the efficiency of verification. Moreover, based on the assumption of bilinear Diffie-Hellman and computational Diffie-Hellman, in the random oracle model, it is proved that the new scheme satisfies the properties of confidentiality and unforgeability.
2015, 37(12): 2845-2851.
doi: 10.11999/JEIT150049
Abstract:
The security of MD-64 block cipher under related-key rectangle attack is studied. Firstly, when the weight of input difference is 1, the differential properties of high order DDOs (Data Dependent Operations) and SPN structures are researched. By the differential properties of high order DDOs and the high probability differential of SPN structures, two related-key differentials are constructed. Secondly, a full round related-key rectangle distinguisher of MD-64 is constructed by connecting two related-key differentials. Thirdly, a related-key rectangle attack is proposed on MD-64, and 32 bits of the master key is recovered with 262 related-key chosen- plain-text, 291.6 encryptions of MD-64 block cipher, and a storage complexity of 266.6 Byte. The success rate of this attack is about 0.961. Analysis results show that MD-64 can not reach the design target under related-key rectangle attack.
The security of MD-64 block cipher under related-key rectangle attack is studied. Firstly, when the weight of input difference is 1, the differential properties of high order DDOs (Data Dependent Operations) and SPN structures are researched. By the differential properties of high order DDOs and the high probability differential of SPN structures, two related-key differentials are constructed. Secondly, a full round related-key rectangle distinguisher of MD-64 is constructed by connecting two related-key differentials. Thirdly, a related-key rectangle attack is proposed on MD-64, and 32 bits of the master key is recovered with 262 related-key chosen- plain-text, 291.6 encryptions of MD-64 block cipher, and a storage complexity of 266.6 Byte. The success rate of this attack is about 0.961. Analysis results show that MD-64 can not reach the design target under related-key rectangle attack.
2015, 37(12): 2852-2857.
doi: 10.11999/JEIT150577
Abstract:
With the rapid development of high-speed railway, the railway mobile communication system has more demands on the handover delay and handover success rate. This paper proposes an advance trigger handover algorithm based on the speed in LTE-R. It is designed to suppress communication interruption and dropped calls caused by that the signaling interaction and setting pre-bearer switching point get rise to handover too early or too late. The theoretical analysis is established in the signaling flow chart, and it is shown that the advance handover algorithm can shorten the handoff latency, which TDD frame structure type two has the most significant time effect. Finally, the simulation shows that compared with the conventional handover algorithm, the advance handover algorithm has a higher and more stable handover success rate. It provides a technical support for the future railway private network applications to LTE-R.
With the rapid development of high-speed railway, the railway mobile communication system has more demands on the handover delay and handover success rate. This paper proposes an advance trigger handover algorithm based on the speed in LTE-R. It is designed to suppress communication interruption and dropped calls caused by that the signaling interaction and setting pre-bearer switching point get rise to handover too early or too late. The theoretical analysis is established in the signaling flow chart, and it is shown that the advance handover algorithm can shorten the handoff latency, which TDD frame structure type two has the most significant time effect. Finally, the simulation shows that compared with the conventional handover algorithm, the advance handover algorithm has a higher and more stable handover success rate. It provides a technical support for the future railway private network applications to LTE-R.
2015, 37(12): 2858-2865.
doi: 10.11999/JEIT150460
Abstract:
Considering the problem of real-time distributed cooperative spectrum detection of cognitive users, a real-time distributed cooperative spectrum detection algorithm based on diffusion strategy is proposed. Global cost function can be approximated by an alternative localized cost that is amenable to distributed optimization. Each individual node optimizes this alternative cost via a steep-descent procedure that relies solely on interaction within the neighborhood of the node. The local estimate value can be calculated via the iteration procedure. A general model for analyzing the mean and variance of the estimates of the diffusion strategy is derived. The formulas of probability of detection, probability of false alarm and detection threshold are derived. Theoretical analysis and experimental results show that the proposed algorithm can effectively solve the problem of real-time detection signal, can quickly learn and adapt to environmental changes. Compared with average consensus strategy and non-real-time diffusion strategy, the average SNR of the proposed algorithm reduces about 6 dB, while the Pfa below 0.01 and Pd reached to 0.9. The diffusion strategy can satisfy the signal detection in very low SNR.
Considering the problem of real-time distributed cooperative spectrum detection of cognitive users, a real-time distributed cooperative spectrum detection algorithm based on diffusion strategy is proposed. Global cost function can be approximated by an alternative localized cost that is amenable to distributed optimization. Each individual node optimizes this alternative cost via a steep-descent procedure that relies solely on interaction within the neighborhood of the node. The local estimate value can be calculated via the iteration procedure. A general model for analyzing the mean and variance of the estimates of the diffusion strategy is derived. The formulas of probability of detection, probability of false alarm and detection threshold are derived. Theoretical analysis and experimental results show that the proposed algorithm can effectively solve the problem of real-time detection signal, can quickly learn and adapt to environmental changes. Compared with average consensus strategy and non-real-time diffusion strategy, the average SNR of the proposed algorithm reduces about 6 dB, while the Pfa below 0.01 and Pd reached to 0.9. The diffusion strategy can satisfy the signal detection in very low SNR.
2015, 37(12): 2866-2871.
doi: 10.11999/JEIT 141496
Abstract:
In order to eliminate the narrowband limit of stochastic resonance system, and to complete the digitized receiving of linear frequency modulation signal with high gain, a method based on the frequency setting of received digital samples is proposed. In this method, the linear frequency modulation is directly sent to the digitized samples flitting, frequency setting process, and is converted to a single frequency signal, which fits in the subsequent cascaded stochastic resonance system. As a result, the noise energy in broadband signal is converted to signal energy. Theory and simulation results show that the algorithm can demodulate the linear frequency modulation signal, and its processing gain is higher than the existing algorithms by about 2 dB.
In order to eliminate the narrowband limit of stochastic resonance system, and to complete the digitized receiving of linear frequency modulation signal with high gain, a method based on the frequency setting of received digital samples is proposed. In this method, the linear frequency modulation is directly sent to the digitized samples flitting, frequency setting process, and is converted to a single frequency signal, which fits in the subsequent cascaded stochastic resonance system. As a result, the noise energy in broadband signal is converted to signal energy. Theory and simulation results show that the algorithm can demodulate the linear frequency modulation signal, and its processing gain is higher than the existing algorithms by about 2 dB.
2015, 37(12): 2872-2876.
doi: 10.11999/JEIT150440
Abstract:
Based on the modeling of transmission rate maximization problem for Orthogonal Frequency Division Multiplexing (OFDM) underwater acoustic communication, the computing amount for water-filling algorithm is reduced reasonably, which realizes rapid assignments for sub-carriers initial energy. To deal with the serious waste of symbol energy existing in the bit assigning algorithm based on fixed threshold, the new greedy algorithm is brought forward, which improves the OFDM transmission rate remarkably with less cost of additional calculation. Results show that the new algorithm behaves favorable properties of transmission rate and Bit Error Rate (BER) even with incomplete channel estimation information suitable for time-variant underwater acoustic channel.
Based on the modeling of transmission rate maximization problem for Orthogonal Frequency Division Multiplexing (OFDM) underwater acoustic communication, the computing amount for water-filling algorithm is reduced reasonably, which realizes rapid assignments for sub-carriers initial energy. To deal with the serious waste of symbol energy existing in the bit assigning algorithm based on fixed threshold, the new greedy algorithm is brought forward, which improves the OFDM transmission rate remarkably with less cost of additional calculation. Results show that the new algorithm behaves favorable properties of transmission rate and Bit Error Rate (BER) even with incomplete channel estimation information suitable for time-variant underwater acoustic channel.
2015, 37(12): 2877-2884.
doi: 10.11999/JEIT150327
Abstract:
The sampling system based on Gabor frames with exponential reproducing windows holds nice performance for short pulses in general cases, but when the frames are highly redundant, the traditional coefficient oriented methods for subspace detection may fail or have large error. Firstly, the signal oriented idea is introduced and the blocked dual Gabor dictionaries are constructed, finishing the block sparse representation. By introducing the blocked dictionaries, the measurement matrix is constructed and the block-coherence restricted by the coherence of the dictionaries is proposed. Consequently, the synthesis model for signal representation is introduced to subspace detection based on Multiple Measurement Vector problem and the Simultaneous Orthogonal Matching Pursuit is proposed based on blocked-closure(SOMPB,F), using for subspace detection. Additionally, the convergence of the algorithm is proved. At last, simulation experiments prove that the new method improves the recovery rate, decreases the channel numbers and enforces the robustness of the sampling system compared with the traditional methods.
The sampling system based on Gabor frames with exponential reproducing windows holds nice performance for short pulses in general cases, but when the frames are highly redundant, the traditional coefficient oriented methods for subspace detection may fail or have large error. Firstly, the signal oriented idea is introduced and the blocked dual Gabor dictionaries are constructed, finishing the block sparse representation. By introducing the blocked dictionaries, the measurement matrix is constructed and the block-coherence restricted by the coherence of the dictionaries is proposed. Consequently, the synthesis model for signal representation is introduced to subspace detection based on Multiple Measurement Vector problem and the Simultaneous Orthogonal Matching Pursuit is proposed based on blocked-closure(SOMPB,F), using for subspace detection. Additionally, the convergence of the algorithm is proved. At last, simulation experiments prove that the new method improves the recovery rate, decreases the channel numbers and enforces the robustness of the sampling system compared with the traditional methods.
2015, 37(12): 2885-2890.
doi: 10.11999/JEIT150340
Abstract:
By analyzing the signal model of the Incoherently Distribute Source (IDS), a sparse representation based parameter estimation method of IDS is presented. Through using the Toeplitz characteristic and the two point approximation model as well as the Jacobi-Anger expansion model of the covariance matrix of the IDS, the angle spread and the central direction angle of the IDS is estimated by adopting two sparse representation problems. Compared with the present method, the proposed method does not need two dimensional searches and has low computational burden. Simulation results show that the proposed method has good parameter estimation performance in the low signal-to-noise ratio and small snapshot number scenario.
By analyzing the signal model of the Incoherently Distribute Source (IDS), a sparse representation based parameter estimation method of IDS is presented. Through using the Toeplitz characteristic and the two point approximation model as well as the Jacobi-Anger expansion model of the covariance matrix of the IDS, the angle spread and the central direction angle of the IDS is estimated by adopting two sparse representation problems. Compared with the present method, the proposed method does not need two dimensional searches and has low computational burden. Simulation results show that the proposed method has good parameter estimation performance in the low signal-to-noise ratio and small snapshot number scenario.
2015, 37(12): 2891-2897.
doi: 10.11999/JEIT150321
Abstract:
As the beamforming of Nonuniform Linear Array (NLA) may occur grating lobes phenomenon, a beamforming method is proposed for working on the NLA with consecutive difference coarray. Firstly, this method analyzes the array optimization of the NLA based on consecutive difference coarray. Additionally, it can be concluded that the consecutive difference coarray is corresponding to the consecutive wavepath difference, which is applied to reconstruct the Toeplitz covariance matrix of the NLA. According to the Least Constraint Mean Variance (LCMV) rule, the reconstructed covariance matrix can directly be used for robust adaptive beamforming. Due to the similarity between the reconstructed covariance matrix and the covariance matrix of Uniform Linear Array (ULA) with the same aperture, the phase ambiguity will not happen and the grating lobes phenomenon will not occur. Extensive simulations show the robust effectiveness of the proposed method.
As the beamforming of Nonuniform Linear Array (NLA) may occur grating lobes phenomenon, a beamforming method is proposed for working on the NLA with consecutive difference coarray. Firstly, this method analyzes the array optimization of the NLA based on consecutive difference coarray. Additionally, it can be concluded that the consecutive difference coarray is corresponding to the consecutive wavepath difference, which is applied to reconstruct the Toeplitz covariance matrix of the NLA. According to the Least Constraint Mean Variance (LCMV) rule, the reconstructed covariance matrix can directly be used for robust adaptive beamforming. Due to the similarity between the reconstructed covariance matrix and the covariance matrix of Uniform Linear Array (ULA) with the same aperture, the phase ambiguity will not happen and the grating lobes phenomenon will not occur. Extensive simulations show the robust effectiveness of the proposed method.
2015, 37(12): 2898-2905.
doi: 10.11999/JEIT150469
Abstract:
Outlier detection, also called anomaly detection, is an important issue in pattern recognition and knowledge discovery. Previous outlier detection methods can not effectively control the false-alarm probability. To solve the problem, a supervised method based on Normalized Residual (NR) is proposed. Using the training patterns, it first calculates the NR value of the query pattern, which is compared with a predefined detection threshold to determine whether the pattern is an outlier. In this paper, the relationship between the threshold and false-alarm probability is theoretically derived, based on which an appropriate threshold can be chosen. In this way, the desired false-alarm probability can be obtained even when few training patterns are available. Simulations and measured data experiments validate the superior performance of the proposed method on outlier detection and false-alarm probability controlling.
Outlier detection, also called anomaly detection, is an important issue in pattern recognition and knowledge discovery. Previous outlier detection methods can not effectively control the false-alarm probability. To solve the problem, a supervised method based on Normalized Residual (NR) is proposed. Using the training patterns, it first calculates the NR value of the query pattern, which is compared with a predefined detection threshold to determine whether the pattern is an outlier. In this paper, the relationship between the threshold and false-alarm probability is theoretically derived, based on which an appropriate threshold can be chosen. In this way, the desired false-alarm probability can be obtained even when few training patterns are available. Simulations and measured data experiments validate the superior performance of the proposed method on outlier detection and false-alarm probability controlling.
2015, 37(12): 2906-2912.
doi: 10.11999/JEIT150319
Abstract:
To overcome the problem that the deficiency of the appearance model and the motion model often leads to low precision in original Multiple Instance Learning (MIL), a target tracking algorithm is proposed based on multiple instance deep learning. In original MIL algorithm, the image is not represented effectively by Haar-like feature. To improve the tracking precision, a stacked denoising autoencoder is used to learn image features and express the image representations obtained effectively. Selected feature vector could not be replaced in the original MIL algorithm, which has difficulty reflecting the changes of the target and the background．Thus, some weakest discriminative feature vector is replaced with new randomly generated feature vector when weak classifiers are selected. It introduces new information to the target model and adapts to the dynamic changes of the target. Aiming at the deficiency of using motion model where the location of the target is likely to appear within a radius in original MIL algorithm, the particle filter estimates objects location to increase the tracking precision. Compared with the original MIL algorithm and other state-of-the-art trackers in the complex environment, the experiments on variant image sequences show that the proposed algorithm raise the tracking accuracy and the robustness.
To overcome the problem that the deficiency of the appearance model and the motion model often leads to low precision in original Multiple Instance Learning (MIL), a target tracking algorithm is proposed based on multiple instance deep learning. In original MIL algorithm, the image is not represented effectively by Haar-like feature. To improve the tracking precision, a stacked denoising autoencoder is used to learn image features and express the image representations obtained effectively. Selected feature vector could not be replaced in the original MIL algorithm, which has difficulty reflecting the changes of the target and the background．Thus, some weakest discriminative feature vector is replaced with new randomly generated feature vector when weak classifiers are selected. It introduces new information to the target model and adapts to the dynamic changes of the target. Aiming at the deficiency of using motion model where the location of the target is likely to appear within a radius in original MIL algorithm, the particle filter estimates objects location to increase the tracking precision. Compared with the original MIL algorithm and other state-of-the-art trackers in the complex environment, the experiments on variant image sequences show that the proposed algorithm raise the tracking accuracy and the robustness.
2015, 37(12): 2913-2920.
doi: 10.11999/JEIT150323
Abstract:
Transfer learning usually focuses on dealing with small training set in target domain by sharing knowledge generated from source ones, in which one main challenge is divergence metric of distributed samples between training and test data. In order to deal with negative transfer problem caused by improper auxiliary sample selections in source domains, this paper presents a modified covariate-shift multi-source ensemble method with transferability criterion. Firstly, transferability metric of auxiliary samples is defined by joint density estimation in accordance with co-variant transfer principles from source to target, so that the coherency of data distributions is verified. After that, whether transfer learning occurs or not should be determined after evaluating transferability metric in different sources to boost accuracy. Finally, experiments on Caltech256 using GIST demonstrate effectiveness and efficiency in the proposed approach and discussions of performance under diverse selections from auxiliary samples and source domains are presented as well. Experimental results show that the proposed method can sufficiently hold back negative transfer for better learnability in transfer style.
Transfer learning usually focuses on dealing with small training set in target domain by sharing knowledge generated from source ones, in which one main challenge is divergence metric of distributed samples between training and test data. In order to deal with negative transfer problem caused by improper auxiliary sample selections in source domains, this paper presents a modified covariate-shift multi-source ensemble method with transferability criterion. Firstly, transferability metric of auxiliary samples is defined by joint density estimation in accordance with co-variant transfer principles from source to target, so that the coherency of data distributions is verified. After that, whether transfer learning occurs or not should be determined after evaluating transferability metric in different sources to boost accuracy. Finally, experiments on Caltech256 using GIST demonstrate effectiveness and efficiency in the proposed approach and discussions of performance under diverse selections from auxiliary samples and source domains are presented as well. Experimental results show that the proposed method can sufficiently hold back negative transfer for better learnability in transfer style.
2015, 37(12): 2921-2928.
doi: 10.11999/JEIT150190
Abstract:
Determining cameras next best view is a difficult issue in visual field. A next best view approach based on depth image of visual object is proposed by using occlusion and contour information in this paper. Firstly, the occlusion detection is accomplished for the depth image of visual object in current view. Secondly, the unknown regions are constructed according to the occlusion detection result of the depth image and the contour of the visual object, and then the unknown regions are modeled with triangulation-like. Thirdly, the midpoint, normal vector and area of each small triangle and other information are utilized to establish the objective function. Finally, the next best view is obtained by optimizing objective function. Experimental results demonstrate that the approach is feasible and effective.
Determining cameras next best view is a difficult issue in visual field. A next best view approach based on depth image of visual object is proposed by using occlusion and contour information in this paper. Firstly, the occlusion detection is accomplished for the depth image of visual object in current view. Secondly, the unknown regions are constructed according to the occlusion detection result of the depth image and the contour of the visual object, and then the unknown regions are modeled with triangulation-like. Thirdly, the midpoint, normal vector and area of each small triangle and other information are utilized to establish the objective function. Finally, the next best view is obtained by optimizing objective function. Experimental results demonstrate that the approach is feasible and effective.
2015, 37(12): 2929-2934.
doi: 10.11999/JEIT150198
Abstract:
In view of the problem of target echoes interfered by strong direct path-wave in bistatic sonar, a high resolution algorithm of direct path interference suppression is proposed when the deployment is known. The method divides the space of N dimension into two orthogonal subspaces under the condition of constraint, the weight matrix of smoothing MVDR (Minimum Variance Distortional Response) algorithm is decomposed into these subspaces, and the optimal weight is obtained by output power minimization in the constraint subspace. The simulation results show that a deeper null appear at the constraint direction compared to conventional beam when the probable direction just is knew, and it yields distortionless response of multi-coherent unknown signals. The algorithm not only suppresses the direct path interference effectively, but also possesses the capability of high resolution for multi-coherent unknown signals.
In view of the problem of target echoes interfered by strong direct path-wave in bistatic sonar, a high resolution algorithm of direct path interference suppression is proposed when the deployment is known. The method divides the space of N dimension into two orthogonal subspaces under the condition of constraint, the weight matrix of smoothing MVDR (Minimum Variance Distortional Response) algorithm is decomposed into these subspaces, and the optimal weight is obtained by output power minimization in the constraint subspace. The simulation results show that a deeper null appear at the constraint direction compared to conventional beam when the probable direction just is knew, and it yields distortionless response of multi-coherent unknown signals. The algorithm not only suppresses the direct path interference effectively, but also possesses the capability of high resolution for multi-coherent unknown signals.
2015, 37(12): 2935-2940.
doi: 10.11999/JEIT150423
Abstract:
A novel wideband signals Direction-Of-Arrival (DOA) estimation method based on sparse representation is proposed. This algorithm can reduce the storage and calculation of the traditional sparse representation methods in wideband signals process, which is caused by the large dimension of base matrix. The over-complete dictionary is constructed by using one-frequency to replace the 2D combination of frequency and angle. The column number of constructed dictionary only equals to that of single-frequency redundant dictionary. The proposed method first adopts focused thought to stack the different frequency data to the reference frequency and founds the redundant dictionary with a single frequency. Then, a sparse recovery model is established to obtain the DOA estimations, which are coming from following the focus process. At the same time, the Singular Value Decomposition (SVD) is used to summarize each frequency to reduce computation burden further. Finally, an automatic selection criterion for the regularization parameter involved in the proposed approach is introduced. The proposed algorithm can effectively distinguish the correlative signals without any decorrelation processing, and it has higher accuracy and detection possibility. The experiment results indicate that the proposed method is effective to estimate the DOA of wideband signals.
A novel wideband signals Direction-Of-Arrival (DOA) estimation method based on sparse representation is proposed. This algorithm can reduce the storage and calculation of the traditional sparse representation methods in wideband signals process, which is caused by the large dimension of base matrix. The over-complete dictionary is constructed by using one-frequency to replace the 2D combination of frequency and angle. The column number of constructed dictionary only equals to that of single-frequency redundant dictionary. The proposed method first adopts focused thought to stack the different frequency data to the reference frequency and founds the redundant dictionary with a single frequency. Then, a sparse recovery model is established to obtain the DOA estimations, which are coming from following the focus process. At the same time, the Singular Value Decomposition (SVD) is used to summarize each frequency to reduce computation burden further. Finally, an automatic selection criterion for the regularization parameter involved in the proposed approach is introduced. The proposed algorithm can effectively distinguish the correlative signals without any decorrelation processing, and it has higher accuracy and detection possibility. The experiment results indicate that the proposed method is effective to estimate the DOA of wideband signals.
2015, 37(12): 2941-2947.
doi: 10.11999/JEIT141624
Abstract:
The traditional sparse ISAR imaging method mainly considers the recovery of coefficients on individual scatters. However, in the practice situation, the target scatters presented by blocks or groups do not emerge on individual. In this case, the usual sparse recover algorithm can not depict the shape of real target, thus, the group-sparse recover approaches are adopted to reconstruct the coefficients of target scatters. The recovery method based on the Bayesian Group-Sparse modeling and Variational inference (VBGS) uses a hierarchical construction of a general signal prior to model the group sparse signals and contain the merit of Sparse Bayesian Learning (SBL) on parameters learning, as a result, it can reconstruct the group sparse signal better than the usual recover algorithm. The VBGS method uses the variational Bayesian inference approach to estimate the parameters of the unknown signal automatically and does not require the parameter-tuning procedures. Considering the sparse group target, this paper combines the Compress Sensing (CS) theory and the VBGS method to reconstruct the ISAR image. The result of experiments show that the proposed method can improve the imaging accuracy compared with traditional algorithm, and can fit to reconstruct the image of ISAR target which has group structure.
The traditional sparse ISAR imaging method mainly considers the recovery of coefficients on individual scatters. However, in the practice situation, the target scatters presented by blocks or groups do not emerge on individual. In this case, the usual sparse recover algorithm can not depict the shape of real target, thus, the group-sparse recover approaches are adopted to reconstruct the coefficients of target scatters. The recovery method based on the Bayesian Group-Sparse modeling and Variational inference (VBGS) uses a hierarchical construction of a general signal prior to model the group sparse signals and contain the merit of Sparse Bayesian Learning (SBL) on parameters learning, as a result, it can reconstruct the group sparse signal better than the usual recover algorithm. The VBGS method uses the variational Bayesian inference approach to estimate the parameters of the unknown signal automatically and does not require the parameter-tuning procedures. Considering the sparse group target, this paper combines the Compress Sensing (CS) theory and the VBGS method to reconstruct the ISAR image. The result of experiments show that the proposed method can improve the imaging accuracy compared with traditional algorithm, and can fit to reconstruct the image of ISAR target which has group structure.
2015, 37(12): 2948-2955.
doi: 10.11999/JEIT150389
Abstract:
The angle tracking loop in airborne radar facing to a maneuvering target plays a vital role in the joint 3D-tracking of range, velocity and angle. This paper analyses the disadvantage of the conventional Kalman filter algorithm employed to track a maneuvering targets angle, which are a low tracking precision and a slow convergence rate of angle tracking error. In order to solve these problems, a novel angle tracking algorithm called Bend Degree Tracking Loop Filter (BDTLF) is put forward to detect the corners in targets angle curve by bend degree detection and adjust the loop noise bandwidth adaptively to control angle tracking loop. The proposed algorithm accelerates the convergence rate in angle tracking loop, lightens the filtering disturbance around targets angle curve corners, and keeps the continuity of filtering performance. The computer simulation results demonstrate that compared with the angle tracking loop using Kalman filtering algorithm, particle filtering algorithm,-- filtering algorithm or a constant coefficient loop filter, this novel method has a more satisfying performance in angle tracking of weakly maneuvering targets.
The angle tracking loop in airborne radar facing to a maneuvering target plays a vital role in the joint 3D-tracking of range, velocity and angle. This paper analyses the disadvantage of the conventional Kalman filter algorithm employed to track a maneuvering targets angle, which are a low tracking precision and a slow convergence rate of angle tracking error. In order to solve these problems, a novel angle tracking algorithm called Bend Degree Tracking Loop Filter (BDTLF) is put forward to detect the corners in targets angle curve by bend degree detection and adjust the loop noise bandwidth adaptively to control angle tracking loop. The proposed algorithm accelerates the convergence rate in angle tracking loop, lightens the filtering disturbance around targets angle curve corners, and keeps the continuity of filtering performance. The computer simulation results demonstrate that compared with the angle tracking loop using Kalman filtering algorithm, particle filtering algorithm,-- filtering algorithm or a constant coefficient loop filter, this novel method has a more satisfying performance in angle tracking of weakly maneuvering targets.
Joint Transmit and Receive Array Position Error Calibration for Bistatic MIMO Radar Based on Clutter
2015, 37(12): 2956-2963.
doi: 10.11999/JEIT150347
Abstract:
The issue of position error estimation for transmit and receive array of a bistatic Multiple-Input Multiple-Output (MIMO) radar is investigated, and an algorithm for the joint estimation based on clutter echo is proposed. The algorithm is based on the criterion of minimizing the reconstruction mean-square error of clutter echo under the restraint ofl1-norm of clutter coefficient. An alternately iterative and convex optimization algorithm is adopted to complete the estimation of clutter scattering coefficients and the position error of both transmit and receive arrays. The simulation results indicate the effectiveness of the proposed algorithm.
The issue of position error estimation for transmit and receive array of a bistatic Multiple-Input Multiple-Output (MIMO) radar is investigated, and an algorithm for the joint estimation based on clutter echo is proposed. The algorithm is based on the criterion of minimizing the reconstruction mean-square error of clutter echo under the restraint ofl1-norm of clutter coefficient. An alternately iterative and convex optimization algorithm is adopted to complete the estimation of clutter scattering coefficients and the position error of both transmit and receive arrays. The simulation results indicate the effectiveness of the proposed algorithm.
2015, 37(12): 2964-2970.
doi: 10.11999/JEIT150299
Abstract:
To reduce estimation error caused by high correlation level of two transmitted waveforms, a novel method is proposed to design a couple of waveforms with low correlation level, named Phase Only Spectral Approximation Algorithm (POSAA). Firstly, the object function is constructed under the rule of minimizing the integrated sidelobe level. Secondly, the object function is derived in frequency domain based on spectral approximation, according to the relationship between correlation sequences and the power spectral density of waveforms. Finally, the object function is optimized by trust region algorithm using its gradient and Hessian matrix. The numerical simulations have demonstrated that the designed waveforms posses a good correlation level, and the error of received target polarization information using this waveform is much less than others.
To reduce estimation error caused by high correlation level of two transmitted waveforms, a novel method is proposed to design a couple of waveforms with low correlation level, named Phase Only Spectral Approximation Algorithm (POSAA). Firstly, the object function is constructed under the rule of minimizing the integrated sidelobe level. Secondly, the object function is derived in frequency domain based on spectral approximation, according to the relationship between correlation sequences and the power spectral density of waveforms. Finally, the object function is optimized by trust region algorithm using its gradient and Hessian matrix. The numerical simulations have demonstrated that the designed waveforms posses a good correlation level, and the error of received target polarization information using this waveform is much less than others.
2015, 37(12): 2971-2976.
doi: 10.11999/JEIT150646
Abstract:
High velocity and Ionosphereic both modulate the phase of low carrier frequency wide-band linearly frequency modulated radar signal , It make the resolution of Inverse SAR (ISAR) image lower. In order to get clean ISAR image, the effect of high velocity and ionosphereic are both must be removed. Firstly, signal model of ionosphereic target with high-velocity are deduced. The high order phase signal parameter estimation method is proposed, using discrete polynomial-phase transform. Motion compensation is done with the estimated values got by this method. Simulation experiments show that the parameters can be estimated right, it can improve the ISAR image deformed by hyper-velocity and ionosphereic.
High velocity and Ionosphereic both modulate the phase of low carrier frequency wide-band linearly frequency modulated radar signal , It make the resolution of Inverse SAR (ISAR) image lower. In order to get clean ISAR image, the effect of high velocity and ionosphereic are both must be removed. Firstly, signal model of ionosphereic target with high-velocity are deduced. The high order phase signal parameter estimation method is proposed, using discrete polynomial-phase transform. Motion compensation is done with the estimated values got by this method. Simulation experiments show that the parameters can be estimated right, it can improve the ISAR image deformed by hyper-velocity and ionosphereic.
2015, 37(12): 2977-2983.
doi: 10.11999/JEIT150442
Abstract:
Range migration is the basic and troublesome problem in moving target detection for wideband radar. To solve this problem, a detection algorithm of moving targets based on joint-sparse recovery is proposed for wideband radar. Firstly, a prewhitening processing is performed to filter the clutter. Then, a jointly row sparse representation of the wideband signals is derived in frequency/slow-time domain，thus the detection problem is solved via joint-sparse recovery. Finally, by using the inverse Fourier transform, the estimation of the targets scenario is achieved. Numerical results are presented to demonstrate the effectiveness of the proposed algorithm.
Range migration is the basic and troublesome problem in moving target detection for wideband radar. To solve this problem, a detection algorithm of moving targets based on joint-sparse recovery is proposed for wideband radar. Firstly, a prewhitening processing is performed to filter the clutter. Then, a jointly row sparse representation of the wideband signals is derived in frequency/slow-time domain，thus the detection problem is solved via joint-sparse recovery. Finally, by using the inverse Fourier transform, the estimation of the targets scenario is achieved. Numerical results are presented to demonstrate the effectiveness of the proposed algorithm.
2015, 37(12): 2984-2990.
doi: 10.11999/JEIT150588
Abstract:
Long integration is often required to detect weak moving target in sea clutter. However, the Doppler frequency spread and amplitude fluctuation in long integration and limited reference cells resulting from spatial non-homogeneity of sea clutter make the traditional adaptive detector work badly. By observing that the Compound Gaussian Distribution (CGD) with Inverse Gamma (IG) texture gives a good fit to sea clutter and the instantaneous frequency is slowly varying, a combined adaptive detector, namely the Combined Adaptive Generalized Likelihood Ratio Test-Linear Threshold Detector (CA-GLRT-LTD), is proposed in the paper, which consists of the product of the maximal response of the adaptive GLRT-LTD in several continuous short integration intervals. Owing to the optimality of the GLRT-LTD for CG clutter with IG texture, the proposed detector obtains better performance than the Combined Adaptive Normalized Matched Filter (CANMF) detector.
Long integration is often required to detect weak moving target in sea clutter. However, the Doppler frequency spread and amplitude fluctuation in long integration and limited reference cells resulting from spatial non-homogeneity of sea clutter make the traditional adaptive detector work badly. By observing that the Compound Gaussian Distribution (CGD) with Inverse Gamma (IG) texture gives a good fit to sea clutter and the instantaneous frequency is slowly varying, a combined adaptive detector, namely the Combined Adaptive Generalized Likelihood Ratio Test-Linear Threshold Detector (CA-GLRT-LTD), is proposed in the paper, which consists of the product of the maximal response of the adaptive GLRT-LTD in several continuous short integration intervals. Owing to the optimality of the GLRT-LTD for CG clutter with IG texture, the proposed detector obtains better performance than the Combined Adaptive Normalized Matched Filter (CANMF) detector.
2015, 37(12): 2991-2999.
doi: 10.11999/JEIT150314
Abstract:
A Non Local Means (NLM) filtering based on Homogeneous Pixels Preselection (NLM-HPP) is proposed to solve the problem of preserving structural feature and polarimetric scattering properties in speckle reduction of Polarimetric SAR (PolSAR) images. Firstly, this method combines statistical property and polarimetric scattering mechanism to select homogeneous pixels in the filtering process. Secondly, the loss function of structure is introduced to improve the accuracy of similarity measure between pixels in NLM method. Finally, it averages the covariance matrices of homogeneous pixels with the weights according to the refined similarity measure, inducing efficient reduction of the speckle in PolSAR images. The implementation results on real PolSAR images, compared with the Refined Lee filter, Scattering-Model-Based speckle filter and two kinds of Non Local Means filter, demonstrate that the proposed method can reduce speckle effectively, and further retain structural information and polarimetric information in PolSAR images.
A Non Local Means (NLM) filtering based on Homogeneous Pixels Preselection (NLM-HPP) is proposed to solve the problem of preserving structural feature and polarimetric scattering properties in speckle reduction of Polarimetric SAR (PolSAR) images. Firstly, this method combines statistical property and polarimetric scattering mechanism to select homogeneous pixels in the filtering process. Secondly, the loss function of structure is introduced to improve the accuracy of similarity measure between pixels in NLM method. Finally, it averages the covariance matrices of homogeneous pixels with the weights according to the refined similarity measure, inducing efficient reduction of the speckle in PolSAR images. The implementation results on real PolSAR images, compared with the Refined Lee filter, Scattering-Model-Based speckle filter and two kinds of Non Local Means filter, demonstrate that the proposed method can reduce speckle effectively, and further retain structural information and polarimetric information in PolSAR images.
2015, 37(12): 3000-3008.
doi: 10.11999/JEIT150480
Abstract:
Compressed Sensing (CS) reconstruction of hyperspectral images driven by spatial-spectral multihypothesis prediction is proposed in order to take full advantage of spatial and spectral correlation of hyperspectral images. The hyperspectral images are grouped into reference band images and non-reference band images, and the reference band images are reconstructed by Smoothed Projected Landweber (SPL) algorithm. For the non-reference band images, the spatial-spectral multihypothesis prediction model is introduced to improve the reconstruction accuracy. Multihypothesis predictions drawn for an image block of non-reference band image are made not only from spatially surrounding image blocks within an initial non-predicted reconstruction of non-reference band image, but also from the corresponding position and neighboring image blocks within the reconstruction of reference band image. The resulting predictions are used to generate residuals in the projection domain, and the residuals are reconstructed to revise the prediction values. The residuals being typically more compressible than the original images and the iterative execution mode lead to improved reconstruction quality. Tikhonov regularization is utilized to solve the weight coefficients of multihypothesis prediction and structural similarity is used as a criterion to decide whether to change the search window size or not. Cross validation is presented to compute the criterion parameter of iteration termination. Experimental results demonstrate that the proposed algorithm outperforms alternative strategies only using spatial correlation or spectral correlation to predict or not employing prediction and the peak signal-to-noise ratio of its reconstructed images is increased by more than 2 dB.
Compressed Sensing (CS) reconstruction of hyperspectral images driven by spatial-spectral multihypothesis prediction is proposed in order to take full advantage of spatial and spectral correlation of hyperspectral images. The hyperspectral images are grouped into reference band images and non-reference band images, and the reference band images are reconstructed by Smoothed Projected Landweber (SPL) algorithm. For the non-reference band images, the spatial-spectral multihypothesis prediction model is introduced to improve the reconstruction accuracy. Multihypothesis predictions drawn for an image block of non-reference band image are made not only from spatially surrounding image blocks within an initial non-predicted reconstruction of non-reference band image, but also from the corresponding position and neighboring image blocks within the reconstruction of reference band image. The resulting predictions are used to generate residuals in the projection domain, and the residuals are reconstructed to revise the prediction values. The residuals being typically more compressible than the original images and the iterative execution mode lead to improved reconstruction quality. Tikhonov regularization is utilized to solve the weight coefficients of multihypothesis prediction and structural similarity is used as a criterion to decide whether to change the search window size or not. Cross validation is presented to compute the criterion parameter of iteration termination. Experimental results demonstrate that the proposed algorithm outperforms alternative strategies only using spatial correlation or spectral correlation to predict or not employing prediction and the peak signal-to-noise ratio of its reconstructed images is increased by more than 2 dB.
2015, 37(12): 3009-3015.
doi: 10.11999/JEIT150047
Abstract:
This paper presents a robust optimal guidance control law for precise three-dimensional (3D) trajectory tracking of an Unmanned Aerial Vehicle (UAV) in wind disturbance. The wind disturbance is considered in the UAVs kinematic model. The reference path is considered as a trajectory of a virtual target. Feedback linearization is used to transform the nonlinear dynamics of the UAV to linear state equations. Based on the assumption that the wind disturbance can be known precisely, an optimal control law is derived for the UAV's 3D trajectory tracking using the LQR (Linear Quadratic Regulator). Then considering the unknown wind disturbance, a robust term is designed to replace the unknown wind disturbance, and a robust optimal control law is obtained. Global asymptotic stability of the closed-loop system is proved by Lyapunov stability theory. Simulations show that the proposed control law can achieve precise 3D UAV trajectory tracking with wind disturbance attenuation, and has good tracking performance.
This paper presents a robust optimal guidance control law for precise three-dimensional (3D) trajectory tracking of an Unmanned Aerial Vehicle (UAV) in wind disturbance. The wind disturbance is considered in the UAVs kinematic model. The reference path is considered as a trajectory of a virtual target. Feedback linearization is used to transform the nonlinear dynamics of the UAV to linear state equations. Based on the assumption that the wind disturbance can be known precisely, an optimal control law is derived for the UAV's 3D trajectory tracking using the LQR (Linear Quadratic Regulator). Then considering the unknown wind disturbance, a robust term is designed to replace the unknown wind disturbance, and a robust optimal control law is obtained. Global asymptotic stability of the closed-loop system is proved by Lyapunov stability theory. Simulations show that the proposed control law can achieve precise 3D UAV trajectory tracking with wind disturbance attenuation, and has good tracking performance.
2015, 37(12): 3016-3024.
doi: 10.11999/JEIT150289
Abstract:
Exact and real-time kinematics model plays a very important role in the mobile robot motion control and path planning. Compared to the off-line model estimation, based on an Instantaneous Centers of Rotation (ICRs) based kinematic model of skid-steering, an Extend Kalman Filter (EKF) method is used to estimate ICRs values on specific terrain on line. Terrains are identified by introducing k-Nearest Neighbors (kNN) algorithm when the robot moves on different terrains. Based on terrain classification, an Adaptive Kalman Filter (AKF) is used to adjust the filter parameters. The simulation and experiment results show that this method can converge very fast and estimate the ICRs value accurately with 3 seconds.
Exact and real-time kinematics model plays a very important role in the mobile robot motion control and path planning. Compared to the off-line model estimation, based on an Instantaneous Centers of Rotation (ICRs) based kinematic model of skid-steering, an Extend Kalman Filter (EKF) method is used to estimate ICRs values on specific terrain on line. Terrains are identified by introducing k-Nearest Neighbors (kNN) algorithm when the robot moves on different terrains. Based on terrain classification, an Adaptive Kalman Filter (AKF) is used to adjust the filter parameters. The simulation and experiment results show that this method can converge very fast and estimate the ICRs value accurately with 3 seconds.
2015, 37(12): 3025-3029.
doi: 10.11999/JEIT150203
Abstract:
The theory of near-field to far-field transformation using spherical-wave expansions is the key to implement the spherical near-field antenna measurement system. It can develop the field in the space which is built by antenna expanding into the sum of spherical wave functions. Because of its complex formula, it will consume a long time to compute. The FFT transformation and the ideas of matrix are put into used in this paper, so the compute speed can be improved and the compute time can be saved. Using this method to testify the near-field data and the far-field data of a horn antenna, the results show that the far-field pattern computed from near-field date and the far-field pattern from theoretical integral equations are compared very well. It is approved that this method can guarantee the calculation precision and shortens the compute time at the same time.
The theory of near-field to far-field transformation using spherical-wave expansions is the key to implement the spherical near-field antenna measurement system. It can develop the field in the space which is built by antenna expanding into the sum of spherical wave functions. Because of its complex formula, it will consume a long time to compute. The FFT transformation and the ideas of matrix are put into used in this paper, so the compute speed can be improved and the compute time can be saved. Using this method to testify the near-field data and the far-field data of a horn antenna, the results show that the far-field pattern computed from near-field date and the far-field pattern from theoretical integral equations are compared very well. It is approved that this method can guarantee the calculation precision and shortens the compute time at the same time.
2015, 37(12): 3030-3040.
doi: 10.11999/JEIT150249
Abstract:
With deep understanding of the characteristics of And-Inverter Cone (AIC), an alternative logic element for FPGA, a series of improvements are proposed to get an optimized interconnect architecture inside the logic cluster. The enhancements include removing the output crossbar, adopting Inverter-Suffixed Crossbar (ISC), optimizing the low load circuit path, dividing the feedback and output function, restricting the output level of AIC and removing the middle crossbar, mixing with the LUT element. An optimized architecture is derived through amounts of experiments. Compared to Stratix IV, Altera, the area of cluster is reduced by 9.06%.Implemented on the new AIC architecture, the average area-delay product of MCNC benchmarks are reduced by 40.82%; the average area-delay product of VTR benchmarks is reduced by 17.38%. Compared to the original AIC-based FPGA architecture, the area of AIC cluster is reduced by 23.16%. Implemented on the new AIC architecture, the average area-delay product of MCNC benchmarks are reduced by 27.15%; the average area-delay product of VTR benchmarks are reduced by 15.26%.
With deep understanding of the characteristics of And-Inverter Cone (AIC), an alternative logic element for FPGA, a series of improvements are proposed to get an optimized interconnect architecture inside the logic cluster. The enhancements include removing the output crossbar, adopting Inverter-Suffixed Crossbar (ISC), optimizing the low load circuit path, dividing the feedback and output function, restricting the output level of AIC and removing the middle crossbar, mixing with the LUT element. An optimized architecture is derived through amounts of experiments. Compared to Stratix IV, Altera, the area of cluster is reduced by 9.06%.Implemented on the new AIC architecture, the average area-delay product of MCNC benchmarks are reduced by 40.82%; the average area-delay product of VTR benchmarks is reduced by 17.38%. Compared to the original AIC-based FPGA architecture, the area of AIC cluster is reduced by 23.16%. Implemented on the new AIC architecture, the average area-delay product of MCNC benchmarks are reduced by 27.15%; the average area-delay product of VTR benchmarks are reduced by 15.26%.
2015, 37(12): 3041-3045.
doi: 10.11999/JEIT150455
Abstract:
In this study, a complex-valued Joint Approximate Diagonalization of Eigen-matrices (JADE)- Independent Component Analysis (JADE-ICA) algorithm is proposed for the Polarization Division Multiplexed in Optical Orthogonal Frequency Division Multiplexing (PDM-OOFDM) systems. Generally, the Constant Modulus Algorithm (CMA) method is used to devise polarization signals in PDM-OOFDM systems. However, this method requires multiple filter coefficients update on CMA, needs more time to converge, and lead it to the polarization multiplexing singularity problem. In this paper, a method based on JADE-ICA algorithm is applied to the PDM-OOFDM systems. With this method, the signals can be separated at the sending and the receiving, which mixed with white Gaussian noise polarization components. Moreover, it improves the separation performance of the system with respect to the polarization signal, while avoiding the traditional CMA polarization multiplexing in the solution of the singularity. Simulations demonstrate the effectiveness of the proposed method to devise signals of polarization in PDM-OOFDM systems.
In this study, a complex-valued Joint Approximate Diagonalization of Eigen-matrices (JADE)- Independent Component Analysis (JADE-ICA) algorithm is proposed for the Polarization Division Multiplexed in Optical Orthogonal Frequency Division Multiplexing (PDM-OOFDM) systems. Generally, the Constant Modulus Algorithm (CMA) method is used to devise polarization signals in PDM-OOFDM systems. However, this method requires multiple filter coefficients update on CMA, needs more time to converge, and lead it to the polarization multiplexing singularity problem. In this paper, a method based on JADE-ICA algorithm is applied to the PDM-OOFDM systems. With this method, the signals can be separated at the sending and the receiving, which mixed with white Gaussian noise polarization components. Moreover, it improves the separation performance of the system with respect to the polarization signal, while avoiding the traditional CMA polarization multiplexing in the solution of the singularity. Simulations demonstrate the effectiveness of the proposed method to devise signals of polarization in PDM-OOFDM systems.
2015, 37(12): 3046-3050.
doi: 10.11999/JEIT150500
Abstract:
To deal with the problem of the serious speckle or composite speckle noise in the image of the strong reflection surface, an image denoising algorithm based on fractional differential enhancement is proposed, which can highlight the granular characteristics of noise, by means of the method of connected region area to remove speckle noise and separate the effective continuous light stripe. Finally, the center of the effective light stripe is extracted with the gray gravity method. By comparison, the method can significantly improve the information entropy and the extraction accuracy of the light stripe center. The fractional differential enhancement algorithm enhance the high frequency information of the image, at the same time it effectively preserves the features of the low frequency information and more details of image texture, and the accuracy of feature extraction is significantly improved.
To deal with the problem of the serious speckle or composite speckle noise in the image of the strong reflection surface, an image denoising algorithm based on fractional differential enhancement is proposed, which can highlight the granular characteristics of noise, by means of the method of connected region area to remove speckle noise and separate the effective continuous light stripe. Finally, the center of the effective light stripe is extracted with the gray gravity method. By comparison, the method can significantly improve the information entropy and the extraction accuracy of the light stripe center. The fractional differential enhancement algorithm enhance the high frequency information of the image, at the same time it effectively preserves the features of the low frequency information and more details of image texture, and the accuracy of feature extraction is significantly improved.
