Email alert
2003 Vol. 25, No. 9
Display Method:
2003, 25(9): 1153-1159.
Abstract:
The structure and algorithm of Priority Ordered Radial Basis Function (POR.BF) Networks is introduced. The concrete training algorithm, calculational methods of likelihood score and verification rule, used for text-independent speaker verification, are proposed. To en-hance the generalization ability, the compressing vectors are applied to construct the inhibitory samples set and three methods including sequential selection, nearest neighbor selection and furthest distance selection are presented for the choose of anti-speakers. Moreover, the out-puts of neurons are weighted by a descendent array. Using these algorithms and methods, the performance is examined by a series of experiments. The results show that under the identi-cal experiment conditions, when the inhibitory set is composed of anti-speakers compressing vectors selected using nearest neighbor method, the Equal Error Rate (EER) using PORBF networks can decreased to 6.83% from 10,56% using conventional VQ method. For speaker verification, the PORBF network provides better performance than the VQ classifier.
The structure and algorithm of Priority Ordered Radial Basis Function (POR.BF) Networks is introduced. The concrete training algorithm, calculational methods of likelihood score and verification rule, used for text-independent speaker verification, are proposed. To en-hance the generalization ability, the compressing vectors are applied to construct the inhibitory samples set and three methods including sequential selection, nearest neighbor selection and furthest distance selection are presented for the choose of anti-speakers. Moreover, the out-puts of neurons are weighted by a descendent array. Using these algorithms and methods, the performance is examined by a series of experiments. The results show that under the identi-cal experiment conditions, when the inhibitory set is composed of anti-speakers compressing vectors selected using nearest neighbor method, the Equal Error Rate (EER) using PORBF networks can decreased to 6.83% from 10,56% using conventional VQ method. For speaker verification, the PORBF network provides better performance than the VQ classifier.
2003, 25(9): 1160-1167.
Abstract:
A complete face recognition system is built based on neural network technologies. Firstly, pure faces are extracted from original face images. Secondly, SOM is utilized to obtain feature reduction. Finally, knowledge-based fuzzy neural networks are used for classification. Experiments on NUSTDBII face database demonstrate the effectiveness of the proposed method and show that it does not sensitive to the variance of pose and expression to a certain extent.
A complete face recognition system is built based on neural network technologies. Firstly, pure faces are extracted from original face images. Secondly, SOM is utilized to obtain feature reduction. Finally, knowledge-based fuzzy neural networks are used for classification. Experiments on NUSTDBII face database demonstrate the effectiveness of the proposed method and show that it does not sensitive to the variance of pose and expression to a certain extent.
2003, 25(9): 1168-1173.
Abstract:
In this paper, the question of Chinese automatic segmentation is discussed using overlaying ambiguity examining method and statistics language model. The multi-time iterative method is applied to train language model, which can produce a better model. The process of training language model is described in detail. The result shows that the perplexity of language model is reduced. The accuracy of segmentation changes with different language model and the reason is analyzed.
In this paper, the question of Chinese automatic segmentation is discussed using overlaying ambiguity examining method and statistics language model. The multi-time iterative method is applied to train language model, which can produce a better model. The process of training language model is described in detail. The result shows that the perplexity of language model is reduced. The accuracy of segmentation changes with different language model and the reason is analyzed.
2003, 25(9): 1174-1179.
Abstract:
Independent Component Analysis (ICA), which is presented in studying the prob-lem of blind source separation, is a novel method of signal processing and data analysis. Using ICA, the source signals can be extracted only from the observations according to the statistic independent characteristics of the input signals, even without the coefficients of the source sig-nals and the transfer channels. In this paper, an image is considered as a mixture of several independent feature causes, and the ICA is used both on the original image and the watermark image respectively, A new approach of embodying digital watermark is proposed. The simu-lation results show that the embedded digital watermark can be extracted correctly, and the robustness exists in the new approach.
Independent Component Analysis (ICA), which is presented in studying the prob-lem of blind source separation, is a novel method of signal processing and data analysis. Using ICA, the source signals can be extracted only from the observations according to the statistic independent characteristics of the input signals, even without the coefficients of the source sig-nals and the transfer channels. In this paper, an image is considered as a mixture of several independent feature causes, and the ICA is used both on the original image and the watermark image respectively, A new approach of embodying digital watermark is proposed. The simu-lation results show that the embedded digital watermark can be extracted correctly, and the robustness exists in the new approach.
2003, 25(9): 1180-1186.
Abstract:
A new blind beamforming algorithm for multitargets is presented in this paper. This method can estimate the weights of beamformers blindly by minimizing a new cost function based on least-squares constant modulus algorithm so as to separate the multitargets and find the direction. At the same time, a developed neural network is used to compute the inverse of a matrix, which not only improve the computing speed but also use easily hardware to carry out the blind beamforming. Simulation proves correctness of this algorithm.
A new blind beamforming algorithm for multitargets is presented in this paper. This method can estimate the weights of beamformers blindly by minimizing a new cost function based on least-squares constant modulus algorithm so as to separate the multitargets and find the direction. At the same time, a developed neural network is used to compute the inverse of a matrix, which not only improve the computing speed but also use easily hardware to carry out the blind beamforming. Simulation proves correctness of this algorithm.
2003, 25(9): 1187-1193.
Abstract:
3D reconstruction of targets in Interferometric SAR (InSAR) refers to determine targets 3D position based on principal measurements of radar and the InSAR imaging geom-etry. Several algorithms have boon proposed for InSAR. These algorithms rely on simplifying assumptions, such as the; flat oarth approximation and the plane wave approximation. Signifi-cant location errors of targets have been introduced by these approximations. In this paper, a new algorithm is proposed on the models of spherical earth and spherical wave. The solution is given in a closed form only using the measured quantities: orbit, parameters, range time, Doppler shift, and interferomotric phase. This analytical solution demonstrates more exact than the existing algorithms.
3D reconstruction of targets in Interferometric SAR (InSAR) refers to determine targets 3D position based on principal measurements of radar and the InSAR imaging geom-etry. Several algorithms have boon proposed for InSAR. These algorithms rely on simplifying assumptions, such as the; flat oarth approximation and the plane wave approximation. Signifi-cant location errors of targets have been introduced by these approximations. In this paper, a new algorithm is proposed on the models of spherical earth and spherical wave. The solution is given in a closed form only using the measured quantities: orbit, parameters, range time, Doppler shift, and interferomotric phase. This analytical solution demonstrates more exact than the existing algorithms.
2003, 25(9): 1194-1199.
Abstract:
Distributed Spaceborne Interferometric SAR(DSI-SAR) can be explored to imple-ment three possibilities of the coherent combination of SAR, images, which are DEM, ocean current imaging and improving resolution in range and azimuth. So the baseline design for DSI-SAR is more complicated than that for INSAR. In this paper, formation flying of small satellites and geometry of DSI-SAR, are discussed and the requirements of three SAR coherent combinations for baselines are analyzed. Furthermore, three performance trade-offs in DSI-SAR baseline design are presented. According to these trade-offs, a method to design baseline is given and applied to an L-band DSI-SAR.
Distributed Spaceborne Interferometric SAR(DSI-SAR) can be explored to imple-ment three possibilities of the coherent combination of SAR, images, which are DEM, ocean current imaging and improving resolution in range and azimuth. So the baseline design for DSI-SAR is more complicated than that for INSAR. In this paper, formation flying of small satellites and geometry of DSI-SAR, are discussed and the requirements of three SAR coherent combinations for baselines are analyzed. Furthermore, three performance trade-offs in DSI-SAR baseline design are presented. According to these trade-offs, a method to design baseline is given and applied to an L-band DSI-SAR.
2003, 25(9): 1200-1205.
Abstract:
In this paper, the intrinsic relationship between registration accuracy and probability of residues referred to in phase unwrapping is investigated. Based on the derived model, it is presented that the number of residues can be used as a criterion to judge the registration results and select the optimal phase fringe image, which will reduce the error caused by misregistration as possible as much. The experiments show the effectiveness of this approach.
In this paper, the intrinsic relationship between registration accuracy and probability of residues referred to in phase unwrapping is investigated. Based on the derived model, it is presented that the number of residues can be used as a criterion to judge the registration results and select the optimal phase fringe image, which will reduce the error caused by misregistration as possible as much. The experiments show the effectiveness of this approach.
2003, 25(9): 1206-1212.
Abstract:
In this paper, a blind channel estimator based on the aid of Pscudo-Pilot-Symbols (PPS) is proposed in terms of the two characters of OFDM system: the finite alphabet property of information symbols and independent property between subchannels. Unlike pilots which are used for estimation of channel and must waste some useful bandwidth, the PPSs, whose power can be boosted 3dB or 6dB to suppress efficiently the additive noise, is useful data transmitted, thus efficiency of the OFDM has been increased. Performance simulation of the proposed estimator including Mean Squares Error (MSE) of channel and uncoded Bit Error Rate(BER) have been taken, and the results show the estimator is efficient and very flexible.
In this paper, a blind channel estimator based on the aid of Pscudo-Pilot-Symbols (PPS) is proposed in terms of the two characters of OFDM system: the finite alphabet property of information symbols and independent property between subchannels. Unlike pilots which are used for estimation of channel and must waste some useful bandwidth, the PPSs, whose power can be boosted 3dB or 6dB to suppress efficiently the additive noise, is useful data transmitted, thus efficiency of the OFDM has been increased. Performance simulation of the proposed estimator including Mean Squares Error (MSE) of channel and uncoded Bit Error Rate(BER) have been taken, and the results show the estimator is efficient and very flexible.
2003, 25(9): 1213-1217.
Abstract:
Estimation accuracy of nonlinear parameter not only relates to the random measurement noise, but also relates to the system observability, provided that system observability is very low and the parameters intermix. Aiming at this, a preprocessing method based on measurements selection is proposed to strengthen the system observability. As an example, the preprocessed subset is used to estimate the system errors of radar. It is shown that this method can improve the estimation accuracy.
Estimation accuracy of nonlinear parameter not only relates to the random measurement noise, but also relates to the system observability, provided that system observability is very low and the parameters intermix. Aiming at this, a preprocessing method based on measurements selection is proposed to strengthen the system observability. As an example, the preprocessed subset is used to estimate the system errors of radar. It is shown that this method can improve the estimation accuracy.
2003, 25(9): 1218-1224.
Abstract:
The technique of the Linear Frequency Modulation(LFM) is proposed to solve the problems of the inter-symbol interference in this paper. For comparing with other modulation techniques, the BER is analyzed, and the simulation results are given. The theoretical prediction and the simulation results indicate that the LFM is very useful to depress the inter-symbol interference. The high transmitting speed is held in some areas where the average of the multi-path time dispersion is larger. The merits of the LFM are that it is not necessary to have an equalizer or the spread spectrum code modulation.
The technique of the Linear Frequency Modulation(LFM) is proposed to solve the problems of the inter-symbol interference in this paper. For comparing with other modulation techniques, the BER is analyzed, and the simulation results are given. The theoretical prediction and the simulation results indicate that the LFM is very useful to depress the inter-symbol interference. The high transmitting speed is held in some areas where the average of the multi-path time dispersion is larger. The merits of the LFM are that it is not necessary to have an equalizer or the spread spectrum code modulation.
2003, 25(9): 1225-1229.
Abstract:
This paper proposes a fast motion estimation algorithm for MPEG-4 shape coding. The proposed algorithm takes advantage of inherent properties of shape coding as well as shape information, i.e., context-based motion estimation feature, motion vector correlation between neighboring binary alpha blocks, and binary feature of shape information. Simulation results show that the algorithm can reduce the computational complexity of motion estimation for shape coding and achieve a high processing speed. Therefore, the algorithm is suitable for real-time software implementation of MPEG-4 shape coding.
This paper proposes a fast motion estimation algorithm for MPEG-4 shape coding. The proposed algorithm takes advantage of inherent properties of shape coding as well as shape information, i.e., context-based motion estimation feature, motion vector correlation between neighboring binary alpha blocks, and binary feature of shape information. Simulation results show that the algorithm can reduce the computational complexity of motion estimation for shape coding and achieve a high processing speed. Therefore, the algorithm is suitable for real-time software implementation of MPEG-4 shape coding.
2003, 25(9): 1230-1236.
Abstract:
The wavelet transform is ideal for image coding due to its capability for multi-resolution representation. This paper presents a remote surveillance system based on the wavelet image coding. The system contains two subsystems, the local monitor and the remote control. The local monitor consists of the image capture module, the wavelet coding module which has two blocks of 2-D wavelet transform and the zerotree coding, and the transmit module. The hardware structure of each module is designed, and the Verilog HDL modules of the local monitor are programmed and simulated. The hardware is then synthesized to FPGA. The receiver program for the remote control is designed, and the remote surveillance system is implemented successfully.
The wavelet transform is ideal for image coding due to its capability for multi-resolution representation. This paper presents a remote surveillance system based on the wavelet image coding. The system contains two subsystems, the local monitor and the remote control. The local monitor consists of the image capture module, the wavelet coding module which has two blocks of 2-D wavelet transform and the zerotree coding, and the transmit module. The hardware structure of each module is designed, and the Verilog HDL modules of the local monitor are programmed and simulated. The hardware is then synthesized to FPGA. The receiver program for the remote control is designed, and the remote surveillance system is implemented successfully.
2003, 25(9): 1237-1242.
Abstract:
In this paper, a new method is proposed for seeking the upper bounds of maximum linear bias for block ciphers, which is especially applicable to a class of Feistel ciphers that key is XORed with data. This technique consists of two steps. Firstly, the mathematical relationship between linear bias of ciphers and linear bias of round function F and S-box respectively is given by carrying out strictly mathematical expression of linear bias for ciphers. Next, the upper bounds of linear bias for ciphers are determined by seeking the solution with minimum weight for linear equation group. Using this method the upper bounds of linear bias within 32 rounds are given.
In this paper, a new method is proposed for seeking the upper bounds of maximum linear bias for block ciphers, which is especially applicable to a class of Feistel ciphers that key is XORed with data. This technique consists of two steps. Firstly, the mathematical relationship between linear bias of ciphers and linear bias of round function F and S-box respectively is given by carrying out strictly mathematical expression of linear bias for ciphers. Next, the upper bounds of linear bias for ciphers are determined by seeking the solution with minimum weight for linear equation group. Using this method the upper bounds of linear bias within 32 rounds are given.
2003, 25(9): 1243-1249.
Abstract:
This paper introduces RSVP, MRSVP and the QoS guarantee mechanism in mobile IP by utilizing the MRSVP. Three HMRSVP schemes are described and compared in detail, especially the HMRSVP with pointer forwarding. A modified one is proposed. In this scheme, MH in a hierarchical region uses pointer forwarding scheme, which can utilize the feature of movement locality. When it moves into a new region, it is required to update the RSVP path and begin the new pointer forwarding process. Also, the feature of locality is discussed thoroughly. The modified scheme is modeled and analyzed in two different cases. The results show that it has better performance.
This paper introduces RSVP, MRSVP and the QoS guarantee mechanism in mobile IP by utilizing the MRSVP. Three HMRSVP schemes are described and compared in detail, especially the HMRSVP with pointer forwarding. A modified one is proposed. In this scheme, MH in a hierarchical region uses pointer forwarding scheme, which can utilize the feature of movement locality. When it moves into a new region, it is required to update the RSVP path and begin the new pointer forwarding process. Also, the feature of locality is discussed thoroughly. The modified scheme is modeled and analyzed in two different cases. The results show that it has better performance.
2003, 25(9): 1250-1254.
Abstract:
Admission control is one of the key policies to guarantee Quality of Service (QoS) in cellular CDMA networks. In this paper, an optimal Call Admisition Control (CAC) scheme based on Semi-Markov Decision Processes (SMDP) is presented to support multiple class ser-vices for QoS wireless networks. Linear programming formulation is used to find the optimal solution, which maximizes the channel utilization while guaranteeing requirements of QoS con-straints. In addition, weighted fair blocking constraints are used to implement flexible QoS requirements. The numerical results show that the performance of this scheme is better than that of threshold based scheme.
Admission control is one of the key policies to guarantee Quality of Service (QoS) in cellular CDMA networks. In this paper, an optimal Call Admisition Control (CAC) scheme based on Semi-Markov Decision Processes (SMDP) is presented to support multiple class ser-vices for QoS wireless networks. Linear programming formulation is used to find the optimal solution, which maximizes the channel utilization while guaranteeing requirements of QoS con-straints. In addition, weighted fair blocking constraints are used to implement flexible QoS requirements. The numerical results show that the performance of this scheme is better than that of threshold based scheme.
2003, 25(9): 1255-1260.
Abstract:
The RCS of metal dihedral corner reflector coated with an anisotropic material is investigated by using the PO method combined with multiple reflections in this paper, which is based on the equivalent currents on an anisotropic material backed by an infinite metal surface illuminated by plane wave. The effect of different material parameters on RCS of the reflector is also discussed. Some useful conclusions are pointed out.
The RCS of metal dihedral corner reflector coated with an anisotropic material is investigated by using the PO method combined with multiple reflections in this paper, which is based on the equivalent currents on an anisotropic material backed by an infinite metal surface illuminated by plane wave. The effect of different material parameters on RCS of the reflector is also discussed. Some useful conclusions are pointed out.
2003, 25(9): 1261-1267.
Abstract:
The lossless constraint conditions of the non-reciprocal passive networks are studied in this paper. It is shown that the impedance matrix [Z] of a non-reciprocal lossless network has the real part, but it is compatible with the law of conservation of energy. In the non-reciprocal case, the positive resistances of the impedance matrix do not represent the. loss of the network, and the negative resistances do not represent the active gain. In addition, the transmission matrix [A] has the characteristic of |det[/l]|=l, which is called as the quasi- reciprocity. The realistic examples are given hi this paper.
The lossless constraint conditions of the non-reciprocal passive networks are studied in this paper. It is shown that the impedance matrix [Z] of a non-reciprocal lossless network has the real part, but it is compatible with the law of conservation of energy. In the non-reciprocal case, the positive resistances of the impedance matrix do not represent the. loss of the network, and the negative resistances do not represent the active gain. In addition, the transmission matrix [A] has the characteristic of |det[/l]|=l, which is called as the quasi- reciprocity. The realistic examples are given hi this paper.
2003, 25(9): 1276-1280.
Abstract:
The characteristics of two parameter CFAR algorithm and K-distribution CFAR algorithm are analyzed. The local window concept of two parameter CFAR is applied to K-distribution CFAR. This method can adapt to the complicated and strong local sea background of SAR imagery and can obtain better results.
The characteristics of two parameter CFAR algorithm and K-distribution CFAR algorithm are analyzed. The local window concept of two parameter CFAR is applied to K-distribution CFAR. This method can adapt to the complicated and strong local sea background of SAR imagery and can obtain better results.
2003, 25(9): 1281-1284.
Abstract:
In this paper, a frame-based spectrum spread watermarking scheme is presented. The watermark is a binary image, and the watermark information are embedded into each frame with their index; at the same time, the correlation of frames is adopted in the scheme. The experimental results show that the embedded watermark is robust against the attacks such as frame drop, swap and collusion. And the extraction of watermark does not require the original video.
In this paper, a frame-based spectrum spread watermarking scheme is presented. The watermark is a binary image, and the watermark information are embedded into each frame with their index; at the same time, the correlation of frames is adopted in the scheme. The experimental results show that the embedded watermark is robust against the attacks such as frame drop, swap and collusion. And the extraction of watermark does not require the original video.
2003, 25(9): 1285-1289.
Abstract:
Polyphase based digital downconvertion is widely used in software electronic system design because it is easy to be realized and also has anti-jamming ability. But this method has blind frequency zone. This paper proposes a new method based on polyphase digital downconvertion which can eliminate the blind frequency zone and realize the receiving of any IF bandpass signal without changing the structure.
Polyphase based digital downconvertion is widely used in software electronic system design because it is easy to be realized and also has anti-jamming ability. But this method has blind frequency zone. This paper proposes a new method based on polyphase digital downconvertion which can eliminate the blind frequency zone and realize the receiving of any IF bandpass signal without changing the structure.
2003, 25(9): 1290-1293.
Abstract:
When the strong tone interference exists in the Direct Sequence Spread Spec-trum(DSSS) system, the traditional linear prediction filter performs poorly. A system model employing the proposed SVD-based technique to reject the single-tone interference in a DSSS system is presented. Simulation results show that SVD method has strong interference sup-pression ability, especially SVD outperforms conventional LMS by a gain of 3dB at BER=10-2. Also the SVD methods avoids the convergence problem that the LMS algorithm has.
When the strong tone interference exists in the Direct Sequence Spread Spec-trum(DSSS) system, the traditional linear prediction filter performs poorly. A system model employing the proposed SVD-based technique to reject the single-tone interference in a DSSS system is presented. Simulation results show that SVD method has strong interference sup-pression ability, especially SVD outperforms conventional LMS by a gain of 3dB at BER=10-2. Also the SVD methods avoids the convergence problem that the LMS algorithm has.
2003, 25(9): 1294-1296.
Abstract:
The electric field summation method of magnetic flux element is proposed to calculate eddy electric field. Based on this method, the time domain formula of eddy electric field generated by plane area rotating in constant magnetic is derived. It also indicates that calculation of eddy electric field using electric field summation method of magnetic flux element is correct.
The electric field summation method of magnetic flux element is proposed to calculate eddy electric field. Based on this method, the time domain formula of eddy electric field generated by plane area rotating in constant magnetic is derived. It also indicates that calculation of eddy electric field using electric field summation method of magnetic flux element is correct.
2003, 25(9): 1268-1275.
Abstract:
In this paper, ATM techniques are introduced and their defects are analyzed, in a view of ATM applications. Advanced Cell Transfer (ACT) technique is one of the Next Generation wideband Networks (NGN) firstly developed in China. Its current progress and its methods to overcome the ATM s defects are introduced and, its future prospected. This paper may help readers to learn in status quo and the comprehensive knowledge of ATM techniques and, understand the trend of network techniques.
In this paper, ATM techniques are introduced and their defects are analyzed, in a view of ATM applications. Advanced Cell Transfer (ACT) technique is one of the Next Generation wideband Networks (NGN) firstly developed in China. Its current progress and its methods to overcome the ATM s defects are introduced and, its future prospected. This paper may help readers to learn in status quo and the comprehensive knowledge of ATM techniques and, understand the trend of network techniques.
