神经网络在辐射源方向信息分选中的应用
|
摘要
辐射源信号分选是雷达对抗中的重要组成部分。面对不断恶化的电子环境,接收到信号特征参数可能发生各种变化,分选的性能急剧恶化,同时分选的实时性受到严峻的考验。然而,辐射源的位置是固定的,信号到达方向(DOA)是信号分选可靠的信息。由于无法确切知道截获的脉冲流类别数目,辐射源方向信息分选本质上属于典型的无监督分类问题。因此,用无监督分类方法研究辐射源方向信息的实时分选具有重要意义。
为了更好研究辐射源方向信息的分选,本文首先利用相关干涉仪测向的方法进行方向信息的提取。在推导了基于圆型阵列的测向公式后,将基线间的和相位信息与差相位信息变换投影到复平面上,对模糊和差相位信息在复平面投影点的分布规律进行了详细的讨论,利用欧式距离的方法解决测向时的模糊问题并对阈值半径等不同因素下解模糊概率进行了仿真与分析。
在获取辐射源方向信息后,选用模糊自谐振神经网络作为无监督分类的方法。在介绍了了模糊自谐振神经网络分选算法后,针对算法存在的问题,并从硬件实现的合理性考虑,对算法的结构进行了调整并提出新的权值更新方法。实验仿真表明,改进后的算法有着更好的分选的性能同时易于硬件逻辑电路实现。
最后,对改进后的模糊自谐振神经网络分选算法进行了FPGA仿真实现。在分析算法并行处理与字长效应的基础上,根据算法结构可划分为输入模块、回忆阶段模块、训练阶段模块,在各模块实现的后用状态机编写同步时序逻辑控制器,使得各个模块间有序执行。通过对顶层模块的仿真实验,表明逻辑电路实现的方向信息分选准确率和利用MATLAB仿真下的准确率接近。
Emtter signal sorting is an important part of radar countermeasure. As the continuous deterioration of electronic environment, receiver possibly receives the forms of radar signal in all kinds of change, and the performance of accuracy sorting deteriorates seriously. At the same time, the performance of real time sorting is subjected to severe test. But the location of radar emitter is fixed, the DOA emitters would be the credible information for radar signal sorting and orientation. However, the intercept and capture of pluses usually lack the number of pluses categories. So radar emitter orientation information sorting belongs to the unsupervised clustering typically. There are important research meanings to research the unknown categories using the direction of arrival emitters for radar signal sorting with unsupervised clustering algorithm.
In order to research the orientation information of emitters sorting in more detail, the paper firstly makes discussion of the method of phrase interferometer to extract orientation information. After derivation direction finding formulas based on the circular array, sum phrase and difference phrase information between baselines is changed and projected in the complex plane. The distribution law of projection points, which are ambiguous sum and difference phrase projection, is also detailed discussed. Then the method of sorting is proposed to solve ambiguous angles problem when direction finding. At the same time, the paper makes simulations and analysis for unwrapping ambiguity probability under different parameter, such as threshold radius.
After acquisition the orientation information of emitters, the Fuzzy Adaptive Resonance Theory Neural Network as an unsupervised clustering algorithm is used in the paper. The Fuzzy ART Neural Network sorting algorithm is introduced. For the drawbacks and unreachable of algorithm when simulation and hardware implementation, the architecture of algorithm is improved. The method of updating weights is also proposed. The better sorting performance of reformulated algorithm is demonstrated by the computer simulation. At the same time, the reformulated algorithm is easy to be implementated with hardware.
Finally, the paper makes FPGA implementation for the reformulated Fuzzy Adaptive Resonance Theory Neural Network algorithm. Based on the foundation of analysis parallel processing and word-length effect, the input module, the recall phrase mudule and the learning phrase mudule are ordered implementated with FPGA according to the architecture of algorithm. Synchronous Sequential Logic controller is made with state machine. Then all the modules can be ordered executed. According to the top layer module simulation, the results demonstrate that the accuracy rate with digital logical hardware is close to the accuracy rate with MATLAB software simulation.
为了更好研究辐射源方向信息的分选,本文首先利用相关干涉仪测向的方法进行方向信息的提取。在推导了基于圆型阵列的测向公式后,将基线间的和相位信息与差相位信息变换投影到复平面上,对模糊和差相位信息在复平面投影点的分布规律进行了详细的讨论,利用欧式距离的方法解决测向时的模糊问题并对阈值半径等不同因素下解模糊概率进行了仿真与分析。
在获取辐射源方向信息后,选用模糊自谐振神经网络作为无监督分类的方法。在介绍了了模糊自谐振神经网络分选算法后,针对算法存在的问题,并从硬件实现的合理性考虑,对算法的结构进行了调整并提出新的权值更新方法。实验仿真表明,改进后的算法有着更好的分选的性能同时易于硬件逻辑电路实现。
最后,对改进后的模糊自谐振神经网络分选算法进行了FPGA仿真实现。在分析算法并行处理与字长效应的基础上,根据算法结构可划分为输入模块、回忆阶段模块、训练阶段模块,在各模块实现的后用状态机编写同步时序逻辑控制器,使得各个模块间有序执行。通过对顶层模块的仿真实验,表明逻辑电路实现的方向信息分选准确率和利用MATLAB仿真下的准确率接近。
Emtter signal sorting is an important part of radar countermeasure. As the continuous deterioration of electronic environment, receiver possibly receives the forms of radar signal in all kinds of change, and the performance of accuracy sorting deteriorates seriously. At the same time, the performance of real time sorting is subjected to severe test. But the location of radar emitter is fixed, the DOA emitters would be the credible information for radar signal sorting and orientation. However, the intercept and capture of pluses usually lack the number of pluses categories. So radar emitter orientation information sorting belongs to the unsupervised clustering typically. There are important research meanings to research the unknown categories using the direction of arrival emitters for radar signal sorting with unsupervised clustering algorithm.
In order to research the orientation information of emitters sorting in more detail, the paper firstly makes discussion of the method of phrase interferometer to extract orientation information. After derivation direction finding formulas based on the circular array, sum phrase and difference phrase information between baselines is changed and projected in the complex plane. The distribution law of projection points, which are ambiguous sum and difference phrase projection, is also detailed discussed. Then the method of sorting is proposed to solve ambiguous angles problem when direction finding. At the same time, the paper makes simulations and analysis for unwrapping ambiguity probability under different parameter, such as threshold radius.
After acquisition the orientation information of emitters, the Fuzzy Adaptive Resonance Theory Neural Network as an unsupervised clustering algorithm is used in the paper. The Fuzzy ART Neural Network sorting algorithm is introduced. For the drawbacks and unreachable of algorithm when simulation and hardware implementation, the architecture of algorithm is improved. The method of updating weights is also proposed. The better sorting performance of reformulated algorithm is demonstrated by the computer simulation. At the same time, the reformulated algorithm is easy to be implementated with hardware.
Finally, the paper makes FPGA implementation for the reformulated Fuzzy Adaptive Resonance Theory Neural Network algorithm. Based on the foundation of analysis parallel processing and word-length effect, the input module, the recall phrase mudule and the learning phrase mudule are ordered implementated with FPGA according to the architecture of algorithm. Synchronous Sequential Logic controller is made with state machine. Then all the modules can be ordered executed. According to the top layer module simulation, the results demonstrate that the accuracy rate with digital logical hardware is close to the accuracy rate with MATLAB software simulation.
引文
[1]李合生,韩宇,蔡英武,等.雷达信号分选关键技术研究综述[J].系统工程与电子技术,2005,17(5):1539~1541.
[2]徐欣,周一宇,卢启中.雷达截获系统实时信号处理技术研究[J].系统工程与电子技术,2001,23(3):12~15.
[3] Mardia H K. New Techiques for the Deinterleaving of Repetitive Sequences[J]. IEE Procceedings,Part F: Radar and Signal Processing,1989,136(4):149-154.
[4]刘旭波,司锡才.雷达信号分选实现的新方法[J].系统工程与电子技术,2010,32(1):53-56.
[5] Yu Zhifu,Ye Fei,Luo Jingqing. A Multi-parameters Synthetic Signal Sorting Algorithm Based on Clustering[C]. ICEMT2007,2007:363-366.
[6]徐欣,冯道旺,周一宇.基于CAM的实时脉冲去交错方法研究[J].电路与系统学报,2001,6(3):94-98.
[7] Kawalec A,Owczarek R. Radar emitter recognition using intrapulse data[C]. Microwaves,Radar and Wireless Communications,2004,MIKON-2004:435–438.
[8] Shieh C S,Lin C T. A Vector Neural Network for Emitter Identification[J]. IEEE Trans. On Antennas and Propagation, 2002,50(8):1120-1127.
[9] Roe J,Cussons S,Feltham A. Knowledge-based Signal Processing for Radar ESM Systems[J]. IEE Proceddings,1990,137(5):293-301.
[10] Ming Gu. Fuzzy ART Neural Network Model and its Application[C]. IEEE International Conference on Intelligent Computing and Intelligent Systems,2009,195-198.
[11] James A,Anderson,Michael T. Radar Signal Categorization Using a Neural Network[J]. Procedings of the IEEE,1990,78(10):1646-1657.
[12] Tang Jinsong,Zhu Zhaoda. A kind of Neural similar to ART Network with Application to Radar Siginal Sorting[C]. Aerospace and Electronics conference,1994,1:398-401.
[13] Cantin M A,Blaquiere Y,Savaria Y, et al. Implementation of the Fuzzy Art Neural Network for Fast Clustering of Radar Pulses[C]. IEEE International Symposium on Circuits and Systems,1998,2:458-461.
[14] Cantin M A,Blaquiere Y,Savaria Y,et al. Analysis of Quantization Effects in a Digital Hardware Implementation of a Fuzzy ART Neural Network Algorithm[C]. IEEE International Symposium on Circuits and Systems,May 28-31,2000,Geneva,Switzerland,3:141-144.
[15] Eric G,Yvon S, Pierre L,et al. A Comparison of Self-Organizing Neural Networks for Fast Clustering of Radar Pulses[J]. ELSEVIER,Signal Processing,1998,64:249-269.
[16] Eric G,Mark A R,Stephen G,et al. Classification of Incomplete Data Using the Fuzzy ARTMAP Neural Network[C]. 2000,4:34-40.
[17] Eric G,Mark A R,Stephen G,et al. A What and Where Fusion Neural Network for Recognition and Tracking of Multiple Radar Emitters [J]. PERGAMON, Neural Networks,2001,14(3):325-344.
[18] Anjaneyulu L,Murthy N S,Sarma N V S N. Radar Emitter Classification Using Self-Organising Neural Network Modles[C]. Proceedings of International Conference on Microwave-08, 2008,431-433.
[19] Sevinc Ilhan,Nevcihan Duru. An Improved Method for Fuzzy Clustering[C]. Fifth International Conference on Soft Computing , Computing with Words and Perceptions in System Analysis,Decision and Control,2009:1-4.
[20]许俊刚,柯有安.自组织神经网络雷达目标识别的研究[J].北京理工大学学报,1992,12(3):67-71.
[21]万建伟,宋小全,皇甫堪等.神经网络在雷达信号分选中的应用.系统工程与电子技术,1995,13(7):29-35.
[22]郭杰,陈文军.一种处理未知雷达信号的聚类方法分选[J].系统工程与电子技术,2006,28(6):583-586.
[23]刘扬,刘璘,杨波.基于二维RBF神经网络的航空雷达信号分选[J].计算机工程与设计,2009,30(1):182-184.
[24]辛化梅,薛林,苏国庆.基于BAM网络和模糊理论的雷达脉冲流模糊联想分选.现代电子技术,2004,13:37-41.
[25]贺洪州,景占荣,徐振华.雷达信号的模糊聚类分选方法[J].航空计算技术,2008,38(5):22-25.
[26]张兴华.模糊聚类分析新算法[J].数学的实践与认识,2005,35(3):138~141.
[27]王永刚.基于模糊聚类在雷达信号分选中的应用[J].电子对抗,2007,113(2):9-12.
[28]刘旭波,司锡才.基于改进的模糊聚类的雷达信号分选[J].弹箭与制导学报,2009,29(5):278-282.
[29] Guo Qiang,Xu Wei,Wang Changhong,et al. Segment Clustering Radar Signal Sorting[C]. International Joint Conference on Computational Sciences and Optimization,2009:943-946.
[30]岳宏伟,罗景青.雷达信号非均匀粒度聚类分选方法[J].火力与指挥控制,2008,33(8):41-44.
[31]张红昌,袁怀林,龚亮亮,等.一种新的未知雷达辐射源聚类分选方法[J].计算机工程与应用,2008,44(27):200-202.
[32]张洪亮,杨文志.基于ISODATA聚类的复杂体制雷达信号分选[J].电子信息对抗技术,2010,25(6):34-38.
[33]向娴,汤建龙.一种基于网格密度聚类的雷达信号分选[J].火控雷达技术,2010,39(4):67-72.
[34]张秀辉,刘已安,曹宁生,等.基于集对分析的雷达信号分选算法[J].现代雷达,2010,32(2):35-37.
[35] Liu Jun , Jim P.Y.Lee Online Clustering Algorithms for Radar Emitter Classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(8):1185-1196.
[36]周红娟,刘帅,金铭,等.基于DOA参数的雷达信号预分选[J].系统工程与电子技术,2009,31(11):2575-2577.
[37] Wang Youyu,Ye Yanfang,Chen Haishan,et al. An Improved Clustering Validity Index for Determining the Number of Malware Clusters[C]. 3rd International Conference on Anti-counterfeiting,Security,and Identification in Communication,2008,544-547.
[38] Nguyen Xuan Vinh,Julien Epps. A Novel Approach for Automatic Number of Clusters Detection in Microarray Data based on Consensus Clustering[C]. 2009Ninth IEEE International Conference on Bioinformatics and Bioengineering,2009,84-91.
[39] Cantin M A,Blaquiere Y,Savaria Y, et al. Implementation of the Fuzzy Art Neural Network for Fast Clustering of Radar Pulses[C]. IEEE International Symposium on Circuits and Systems,1998,2:458-461.
[40] Cantin M A,Blaquiere Y,Savaria Y,et al. Analysis of Quantization Effects in a Digital Hardware Implementation of a Fuzzy ART Neural Network Algorithm[C]. IEEE International Symposium on Circuits and Systems,May 28-31,2000,Geneva,Switzerland,3:141-144.
[41] Granger E,Blaquire Y,Savaria Y, Cantin M A,et al. A VLSI Architecture for Fast Clustering with Fuzzy ART Neural Networks[C]. International Workshop on Neural Networks for Identification , Control, Robotics, and Signal/Image Processing,1996,117-125.
[42] Sudhir G,Akojwar,Rajendra M,Patrikar. Neural Networks based Real TimeClassifier for Wireless Sensor Networks and Framework for VLSI Implementation[C]. IEEE International Symposium on Industrial Electronics,2007,1381-1386.
[43] Chen B,Hoberock L. A Fuzzy Neural Network Architecture for Fuzzy Control and Classification[C]. IEEE International Conference on Neural Networks,1996,2: 1168-1173.
[44]孔超,李占才,王泌,等.一种基于FPGA的SOM神经网络的并行实现[J].计算机工程,2007,33(19):236-240.
[2]徐欣,周一宇,卢启中.雷达截获系统实时信号处理技术研究[J].系统工程与电子技术,2001,23(3):12~15.
[3] Mardia H K. New Techiques for the Deinterleaving of Repetitive Sequences[J]. IEE Procceedings,Part F: Radar and Signal Processing,1989,136(4):149-154.
[4]刘旭波,司锡才.雷达信号分选实现的新方法[J].系统工程与电子技术,2010,32(1):53-56.
[5] Yu Zhifu,Ye Fei,Luo Jingqing. A Multi-parameters Synthetic Signal Sorting Algorithm Based on Clustering[C]. ICEMT2007,2007:363-366.
[6]徐欣,冯道旺,周一宇.基于CAM的实时脉冲去交错方法研究[J].电路与系统学报,2001,6(3):94-98.
[7] Kawalec A,Owczarek R. Radar emitter recognition using intrapulse data[C]. Microwaves,Radar and Wireless Communications,2004,MIKON-2004:435–438.
[8] Shieh C S,Lin C T. A Vector Neural Network for Emitter Identification[J]. IEEE Trans. On Antennas and Propagation, 2002,50(8):1120-1127.
[9] Roe J,Cussons S,Feltham A. Knowledge-based Signal Processing for Radar ESM Systems[J]. IEE Proceddings,1990,137(5):293-301.
[10] Ming Gu. Fuzzy ART Neural Network Model and its Application[C]. IEEE International Conference on Intelligent Computing and Intelligent Systems,2009,195-198.
[11] James A,Anderson,Michael T. Radar Signal Categorization Using a Neural Network[J]. Procedings of the IEEE,1990,78(10):1646-1657.
[12] Tang Jinsong,Zhu Zhaoda. A kind of Neural similar to ART Network with Application to Radar Siginal Sorting[C]. Aerospace and Electronics conference,1994,1:398-401.
[13] Cantin M A,Blaquiere Y,Savaria Y, et al. Implementation of the Fuzzy Art Neural Network for Fast Clustering of Radar Pulses[C]. IEEE International Symposium on Circuits and Systems,1998,2:458-461.
[14] Cantin M A,Blaquiere Y,Savaria Y,et al. Analysis of Quantization Effects in a Digital Hardware Implementation of a Fuzzy ART Neural Network Algorithm[C]. IEEE International Symposium on Circuits and Systems,May 28-31,2000,Geneva,Switzerland,3:141-144.
[15] Eric G,Yvon S, Pierre L,et al. A Comparison of Self-Organizing Neural Networks for Fast Clustering of Radar Pulses[J]. ELSEVIER,Signal Processing,1998,64:249-269.
[16] Eric G,Mark A R,Stephen G,et al. Classification of Incomplete Data Using the Fuzzy ARTMAP Neural Network[C]. 2000,4:34-40.
[17] Eric G,Mark A R,Stephen G,et al. A What and Where Fusion Neural Network for Recognition and Tracking of Multiple Radar Emitters [J]. PERGAMON, Neural Networks,2001,14(3):325-344.
[18] Anjaneyulu L,Murthy N S,Sarma N V S N. Radar Emitter Classification Using Self-Organising Neural Network Modles[C]. Proceedings of International Conference on Microwave-08, 2008,431-433.
[19] Sevinc Ilhan,Nevcihan Duru. An Improved Method for Fuzzy Clustering[C]. Fifth International Conference on Soft Computing , Computing with Words and Perceptions in System Analysis,Decision and Control,2009:1-4.
[20]许俊刚,柯有安.自组织神经网络雷达目标识别的研究[J].北京理工大学学报,1992,12(3):67-71.
[21]万建伟,宋小全,皇甫堪等.神经网络在雷达信号分选中的应用.系统工程与电子技术,1995,13(7):29-35.
[22]郭杰,陈文军.一种处理未知雷达信号的聚类方法分选[J].系统工程与电子技术,2006,28(6):583-586.
[23]刘扬,刘璘,杨波.基于二维RBF神经网络的航空雷达信号分选[J].计算机工程与设计,2009,30(1):182-184.
[24]辛化梅,薛林,苏国庆.基于BAM网络和模糊理论的雷达脉冲流模糊联想分选.现代电子技术,2004,13:37-41.
[25]贺洪州,景占荣,徐振华.雷达信号的模糊聚类分选方法[J].航空计算技术,2008,38(5):22-25.
[26]张兴华.模糊聚类分析新算法[J].数学的实践与认识,2005,35(3):138~141.
[27]王永刚.基于模糊聚类在雷达信号分选中的应用[J].电子对抗,2007,113(2):9-12.
[28]刘旭波,司锡才.基于改进的模糊聚类的雷达信号分选[J].弹箭与制导学报,2009,29(5):278-282.
[29] Guo Qiang,Xu Wei,Wang Changhong,et al. Segment Clustering Radar Signal Sorting[C]. International Joint Conference on Computational Sciences and Optimization,2009:943-946.
[30]岳宏伟,罗景青.雷达信号非均匀粒度聚类分选方法[J].火力与指挥控制,2008,33(8):41-44.
[31]张红昌,袁怀林,龚亮亮,等.一种新的未知雷达辐射源聚类分选方法[J].计算机工程与应用,2008,44(27):200-202.
[32]张洪亮,杨文志.基于ISODATA聚类的复杂体制雷达信号分选[J].电子信息对抗技术,2010,25(6):34-38.
[33]向娴,汤建龙.一种基于网格密度聚类的雷达信号分选[J].火控雷达技术,2010,39(4):67-72.
[34]张秀辉,刘已安,曹宁生,等.基于集对分析的雷达信号分选算法[J].现代雷达,2010,32(2):35-37.
[35] Liu Jun , Jim P.Y.Lee Online Clustering Algorithms for Radar Emitter Classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(8):1185-1196.
[36]周红娟,刘帅,金铭,等.基于DOA参数的雷达信号预分选[J].系统工程与电子技术,2009,31(11):2575-2577.
[37] Wang Youyu,Ye Yanfang,Chen Haishan,et al. An Improved Clustering Validity Index for Determining the Number of Malware Clusters[C]. 3rd International Conference on Anti-counterfeiting,Security,and Identification in Communication,2008,544-547.
[38] Nguyen Xuan Vinh,Julien Epps. A Novel Approach for Automatic Number of Clusters Detection in Microarray Data based on Consensus Clustering[C]. 2009Ninth IEEE International Conference on Bioinformatics and Bioengineering,2009,84-91.
[39] Cantin M A,Blaquiere Y,Savaria Y, et al. Implementation of the Fuzzy Art Neural Network for Fast Clustering of Radar Pulses[C]. IEEE International Symposium on Circuits and Systems,1998,2:458-461.
[40] Cantin M A,Blaquiere Y,Savaria Y,et al. Analysis of Quantization Effects in a Digital Hardware Implementation of a Fuzzy ART Neural Network Algorithm[C]. IEEE International Symposium on Circuits and Systems,May 28-31,2000,Geneva,Switzerland,3:141-144.
[41] Granger E,Blaquire Y,Savaria Y, Cantin M A,et al. A VLSI Architecture for Fast Clustering with Fuzzy ART Neural Networks[C]. International Workshop on Neural Networks for Identification , Control, Robotics, and Signal/Image Processing,1996,117-125.
[42] Sudhir G,Akojwar,Rajendra M,Patrikar. Neural Networks based Real TimeClassifier for Wireless Sensor Networks and Framework for VLSI Implementation[C]. IEEE International Symposium on Industrial Electronics,2007,1381-1386.
[43] Chen B,Hoberock L. A Fuzzy Neural Network Architecture for Fuzzy Control and Classification[C]. IEEE International Conference on Neural Networks,1996,2: 1168-1173.
[44]孔超,李占才,王泌,等.一种基于FPGA的SOM神经网络的并行实现[J].计算机工程,2007,33(19):236-240.