基于小波分析与神经网络的无线信号分类方法的研究
|
摘要
随着通信技术的发展,无线通信环境日益复杂,通信信号在很宽的频带上采用各种调制方式。如何有效的监视和识别这些信号是通信信号处理的一个重要研究课题,是电子对抗的一个重要内容。其广泛应用于信号确认、干扰识别、无线电侦听和信号监测以及软件无线电、卫星通信等领域。
小波变换具有良好的时频局部特性,非常适合分析突变信号和非平稳信号,而神经网络具有自学习、自适应、鲁棒性、容错性等优点,利用神经网络可以提高识别的自动化和智能化。因此,本课题采用小波分析与神经网络等现代信号处理技术,实现对AM、DSB、USB、LSB、FM、PM、2ASK、4ASK、2FSK、4FSK、2PSK、4PSK等12种常用无线信号的分类。
首先,研究了通信信号的调制原理和特点,并用MATLAB实现了这些通信调制信号。其次,设计了一种基于小波分析和神经网络的无线信号调制方式分类的方法。该方法分两步实现。第一步是特征提取,采用Daubechies小波对信号进行七层分解和重构,研究了以各层信号的均方差作为信号的特征矢量的方法。第二步是调制方式分类,设计了用于实现调制信号分类的RBF神经网络,用提取的样本集的特征对RBF神经网络进行训练,用训练好的网络对测试集的信号进行分类。最后,基于MATLAB7.1平台对调制信号分类方法进行仿真试验,实现了十二种调制信号在不同信噪比下的分类。当SNR=5dB时,信号的平均识别率达到了98.58%,最低识别率为97%。仿真结果表明,小波分析和神经网络相结合,可以很好的实现无线信号的分类。
With the development of communication technology, wireless communication has become more and more complicated. Various modulation types are used in the communication signals with broad bandwidth. How to monitor and identify these signals is an essential subject of communication signal processing, and an important part in electronic countermeasures too. At the same time, signal identification is a rapidly evolving area of signal analysis. The automatic identification of signal modulations has been applied many fields,such as identification, interference identification, radio interception and monitoring, satellite communications, etc.
Wavelet transformation has a good localization characteristic in time-frequency domain, while the neural network has characteristics of self-study, self-adaptation, high stabilization and error acceptability. Using neural network can improve the automatization and intelligence of recognition. Based on above, wavelet analysis and neural network are used to implement the classification of the modulation signals. AM、DSB、USB、LSB、FM、PM、2ASK、4ASK、2FSK、4FSK、2PSK、4PSK are researched in this thesis.
The theory of signal modulation is introduced first, and these modulation signals are realized with MATLAB. Then a method of classification based on wavelet neural network is designed in this thesis. In the step of characteristic extraction, wavelet transformation is used to analyze the twelve modulation signals and extract the characteristic parameters. In the step of classification, the characteristics of signals which are extracted in samples are used to train the RBF neural network. RBF export the classification result when the error meets the requirement. Finally the method is simulated with MATALB. Under a 5dB of signal to noise ratio, the average recognition ratio is 98.58%, and the lowest recognition ratio is 97%. The simulation results indicate that the presented method performs well.
小波变换具有良好的时频局部特性,非常适合分析突变信号和非平稳信号,而神经网络具有自学习、自适应、鲁棒性、容错性等优点,利用神经网络可以提高识别的自动化和智能化。因此,本课题采用小波分析与神经网络等现代信号处理技术,实现对AM、DSB、USB、LSB、FM、PM、2ASK、4ASK、2FSK、4FSK、2PSK、4PSK等12种常用无线信号的分类。
首先,研究了通信信号的调制原理和特点,并用MATLAB实现了这些通信调制信号。其次,设计了一种基于小波分析和神经网络的无线信号调制方式分类的方法。该方法分两步实现。第一步是特征提取,采用Daubechies小波对信号进行七层分解和重构,研究了以各层信号的均方差作为信号的特征矢量的方法。第二步是调制方式分类,设计了用于实现调制信号分类的RBF神经网络,用提取的样本集的特征对RBF神经网络进行训练,用训练好的网络对测试集的信号进行分类。最后,基于MATLAB7.1平台对调制信号分类方法进行仿真试验,实现了十二种调制信号在不同信噪比下的分类。当SNR=5dB时,信号的平均识别率达到了98.58%,最低识别率为97%。仿真结果表明,小波分析和神经网络相结合,可以很好的实现无线信号的分类。
With the development of communication technology, wireless communication has become more and more complicated. Various modulation types are used in the communication signals with broad bandwidth. How to monitor and identify these signals is an essential subject of communication signal processing, and an important part in electronic countermeasures too. At the same time, signal identification is a rapidly evolving area of signal analysis. The automatic identification of signal modulations has been applied many fields,such as identification, interference identification, radio interception and monitoring, satellite communications, etc.
Wavelet transformation has a good localization characteristic in time-frequency domain, while the neural network has characteristics of self-study, self-adaptation, high stabilization and error acceptability. Using neural network can improve the automatization and intelligence of recognition. Based on above, wavelet analysis and neural network are used to implement the classification of the modulation signals. AM、DSB、USB、LSB、FM、PM、2ASK、4ASK、2FSK、4FSK、2PSK、4PSK are researched in this thesis.
The theory of signal modulation is introduced first, and these modulation signals are realized with MATLAB. Then a method of classification based on wavelet neural network is designed in this thesis. In the step of characteristic extraction, wavelet transformation is used to analyze the twelve modulation signals and extract the characteristic parameters. In the step of classification, the characteristics of signals which are extracted in samples are used to train the RBF neural network. RBF export the classification result when the error meets the requirement. Finally the method is simulated with MATALB. Under a 5dB of signal to noise ratio, the average recognition ratio is 98.58%, and the lowest recognition ratio is 97%. The simulation results indicate that the presented method performs well.
引文
[1] Weaver C S, Cole C A, Krumland R B, et al. “The automatic classification of modulation types by pattern recognition” AD691069, April, 1969
[2] Polydoros A, Kim K. On the detection and classification of quadrature digital modulations in broad-band noise. IEEE Trans. Communications. 1990, 38(8):1199-1211
[3] Chugg K M, long Chu-Sieng, Polydoros A. Combined likelihood power estimation and multiple hypothesis modulation classification. Signals, Systems and Computers, Conference, Record of the Twenty-Ninth Asilomar. Vo1.2,1996: 1137-1141
[4] Sills J A, Maximum-likelihood modulation classification for PSK/QAM, in: IEEE MILCOM ’99 Proc., vol. 1, 1999: 217–220
[5] Wei Wen, Mendel, J M. Maximum-likelihood classification for digital amplitude- phase modulations[J].IEEE Trans.on Communications. 2000,48(2):189-193
[6] Hsue S Z, Soliman S S,Automatic modulation recognition of digitally modulated signals. MILCOM'89, 37.4.1-37.4.5
[7] Hsue S Z, Soliman S S. Automatic modulation classification using zero crossing. Radar and Signal Processing,IEE Proceedings.1990, 137(6): 459-464
[8] Assaleh K, Farrell K, Mammone R J. A new method of modulation classification for digitally modulated signals. in Proc.IEEE MILCOM, ,1992: 712-716
[9] Lu Mingquan, Xiao Xianci, and Li lemin. Ar modeling based features extraction for multiple signals for modulation recognition, Signal Processing Proceedings, 4th International Conference,1998, vol.2:1384-1388
[10] 戴威,王有政,王京.基于 AR 模型的调制盲识别方法.电子学报.2Q01,Vo1.29, No.12A:1890-1892
[11] Ho K C, Prokopiw W, and Chan Y T. Modulation identification by the wavelet transform, in Proc. IEEE MILCOM,1995:886-890
[12] Ta N P.A wavelet packet approach to radio signal modulation classification.ICCS'94, Singapore, Vol.l 1994: 210-214
[13] Liang Hong, Ho K C. Identification of digital modulation types using the wavelet transform. MILCOM'99, Vol.l: 427-431
[14] Sangwoo Cho, Chong Hyun Lee, Jooh wan chun. Classification of digital modulations using the LPC. National Aerospace and Electronics Conference, Proceedings of the IEEE, 2000, 774-778
[15] Ho K C, Prokopiw, Chan Y T. Modulation identification of digital signals by the wavelet transform. IEEE. Proc. Radar. Sonar Nacig.2000.147:69-176
[16] Samir S, Soliman, Shue-Zen Hsue. Signal classification using statistical moinents. IEEE Traps. Communication.1992,vo1.40: 908-916
[17] Dai Wei, Wang Youzheng, Wang Jing. Joint Power Estimation and Modulation Classification Using Second-and Higher Statistics. Wireless Communications and networking Conference, 2002,IEEE, Vo1.1:155-158
[18] 高永强,陈建安,基于高阶累量的数字调制方式识别,无线通信技术,2006,15(1):26-29
[19] Gardener W A. Spectral Correlation of Modulated Signals: PART I –Analog Modulation. IEE Traps. Communication.1987, Vo1.35: 584-594
[20] Gardener W A. Spectral Correlation of Modulated Signals: PART II –Digital Modulation. IEEE Traps. Communication.1987, Vo1.35: 595-601
[21] Gardener W A, Spooner C W. Cyclic spectral analysis for signal detection and modulation recognition. MILCOM'98, Vol.2: 419-424
[22] 韩国栋 蔡斌 邬江兴.调制分析与识别的谱相关方法. 系统工程与电子技术。2001,23(3):34-36
[23] 曹志刚.钱亚生.现代通信原理.北京:清华大学出版社,1992
[24] 王生兵.无线电信号的调制识别研究:[硕士学位论文].东南大学,2006
[25] 飞思科技产品研发中心,小波分析理论与 MATLAB7 实现,北京:电子工业出版社,2005.
[26] 张晓文,杨煜普,许晓鸣.基于小波变换的特征构造与选择.计算机工程与应用.2003,19: 25-28
[27] Learned R E, Willsky A S. A wavelet packet approach to transient signal classification[J]. Applied and Computational Harmaric Analysis, 1995: 2(3): 256-278
[28] Mallat S, Zhong S. Characterization of signals from multiscale edges[J]. IEEE Trans Pattern Recognition and Machine Intelligence,1992: 14(7):710-732
[29] Senhadji L, Carrault G, Ballanger J J, et al. Comparing wavelet transforms for recognizing cardiac patterns[J].IEEE Engineering in Mechine and Biology,1995:167-172
[30] Hazarika N, Chen J Z, Tsoi A C. Classification of EEG Signals Using the Wavelet Transform. IEEE DSP,1997:89-92
[31] Pittner S, Kamarthi S V. Feature From Wavelet Coefficients for Pattern recognition Tasks[J].IEEE Trans on Pattern Analysis and Machine Intalligence,1999:21(1):83-88
[32] Rioul O, Vetterli M. Wavelets and Signal Processing[J].ASSP Magazine,1991:14-38
[33] Shensa M J. The Discrete Wavelet Transform: Wedding the A-Trous and Mallat Algorithms[J]. IEEE Trans on Signal Processing,1992;40(10):2464-2478
[34] 胡广书.现代信号处理教程. 北京:清华大学出版社,2004
[35] 张贤达.现代信号处理.清华大学出版社(第二版),2002:378-428
[36] Avci, E., Turkoglu, I., Poyraz, M. (2005a). Intelligent target recognition based on wavelet packet neural network. Experts Systems with Applications, 29(1)
[37] Avci, E., Turkoglu, I., Poyraz, M. (2005b). In Intelligent target recognition based on wavelet adaptive network based fuzzy inference system. Lecture notes in computer science. Berlin: Springer-Verlag, May. 2005(3522):594–601
[38] Avci E. An expert Discrete Wavelet Adaptive Network Based Fuzzy Inference System for digital modulation recognition. Expert Systems with Applications.2007, 33(3),582-589
[39] Avci E. Performance comparison of wavelet families for analog modulation classification using expert discrete wavelet neural network system .Expert Systems with Applications. 2007, 33(1),23-35
[40] 夏克文,智能信息处理.天津:河北工业大学,2004
[41] 焦李成,神经网络系统理论[M],西安:电子科技大学出版社,1992
[42] 靳蕃 编著,神经计算智能基础,成都:西南交通大学出版社,2000 年
[43] 徐秉铮,张百灵,韦岗.神经网络理论与应用.华南理工大学出版社,1994
[44] Powell M J D. Radial basis functions for multivariate interpolation. In: Mason J M, Cox M, eds. Algorithms for Approximation.1985, 143-167
[45] Broomhead D S and D Lowe. Multivariate functional interpolation and adaptive networks. Complex Systems,1988(2):321-355
[46] 夏克文,宋建平.应用带有非线性连接权的神经网络识别水泥胶结质量.西安交通大学学报. 2003,37(2):192-195
[47] 飞思科技产品研发中心,神经网络理论与 MATLAB7 实现.北京:电子工业出版社,2005.
[48] 罗利春.无线电侦察信号分析与处理.国防工业出版社,2003:81-85
[49] Blake R.现代通信系统.电子工业出版社,2003:371-394
[50] Wong M L D, Nandi A K. Automatic digital modulation recognition using artificial neural network and genetic algorithm. Signal Processing, 2004,84:351–365
[51] Wu Z, Ren G, Wang X, et al. Automatic digital modulation recognition using wavelet transform and neural networks. ISNN 2004, LNCS 3173 :936–940
[52] Wu Z, Wang X, Liu C, et al. Automatic digital modulation recognition based on ART2A-DWNN. ISNN 2005, LNCS 3497 :381–386
[53] Zhang Q, Benveniste A. Wavelet network .IEEE Transactions on Neural Networks. 3(6):889–898
[54] 李建新,刘乃安,刘继平.现代通信系统分析与仿真-MATLAB通信工具箱.西安:西安电子科技大学出版社,2000.154-168
[55] Zhang J, Walter G G, Miao Y, et al. Wavelet neural networks for function learning. IEEE Transactions on Signal Processing, 43(6)
[56] Musavi M T, Ahmed W, Chan K H, et al. On the training of radial basis function classifiers, Neural Networks,1992. 595-603
[2] Polydoros A, Kim K. On the detection and classification of quadrature digital modulations in broad-band noise. IEEE Trans. Communications. 1990, 38(8):1199-1211
[3] Chugg K M, long Chu-Sieng, Polydoros A. Combined likelihood power estimation and multiple hypothesis modulation classification. Signals, Systems and Computers, Conference, Record of the Twenty-Ninth Asilomar. Vo1.2,1996: 1137-1141
[4] Sills J A, Maximum-likelihood modulation classification for PSK/QAM, in: IEEE MILCOM ’99 Proc., vol. 1, 1999: 217–220
[5] Wei Wen, Mendel, J M. Maximum-likelihood classification for digital amplitude- phase modulations[J].IEEE Trans.on Communications. 2000,48(2):189-193
[6] Hsue S Z, Soliman S S,Automatic modulation recognition of digitally modulated signals. MILCOM'89, 37.4.1-37.4.5
[7] Hsue S Z, Soliman S S. Automatic modulation classification using zero crossing. Radar and Signal Processing,IEE Proceedings.1990, 137(6): 459-464
[8] Assaleh K, Farrell K, Mammone R J. A new method of modulation classification for digitally modulated signals. in Proc.IEEE MILCOM, ,1992: 712-716
[9] Lu Mingquan, Xiao Xianci, and Li lemin. Ar modeling based features extraction for multiple signals for modulation recognition, Signal Processing Proceedings, 4th International Conference,1998, vol.2:1384-1388
[10] 戴威,王有政,王京.基于 AR 模型的调制盲识别方法.电子学报.2Q01,Vo1.29, No.12A:1890-1892
[11] Ho K C, Prokopiw W, and Chan Y T. Modulation identification by the wavelet transform, in Proc. IEEE MILCOM,1995:886-890
[12] Ta N P.A wavelet packet approach to radio signal modulation classification.ICCS'94, Singapore, Vol.l 1994: 210-214
[13] Liang Hong, Ho K C. Identification of digital modulation types using the wavelet transform. MILCOM'99, Vol.l: 427-431
[14] Sangwoo Cho, Chong Hyun Lee, Jooh wan chun. Classification of digital modulations using the LPC. National Aerospace and Electronics Conference, Proceedings of the IEEE, 2000, 774-778
[15] Ho K C, Prokopiw, Chan Y T. Modulation identification of digital signals by the wavelet transform. IEEE. Proc. Radar. Sonar Nacig.2000.147:69-176
[16] Samir S, Soliman, Shue-Zen Hsue. Signal classification using statistical moinents. IEEE Traps. Communication.1992,vo1.40: 908-916
[17] Dai Wei, Wang Youzheng, Wang Jing. Joint Power Estimation and Modulation Classification Using Second-and Higher Statistics. Wireless Communications and networking Conference, 2002,IEEE, Vo1.1:155-158
[18] 高永强,陈建安,基于高阶累量的数字调制方式识别,无线通信技术,2006,15(1):26-29
[19] Gardener W A. Spectral Correlation of Modulated Signals: PART I –Analog Modulation. IEE Traps. Communication.1987, Vo1.35: 584-594
[20] Gardener W A. Spectral Correlation of Modulated Signals: PART II –Digital Modulation. IEEE Traps. Communication.1987, Vo1.35: 595-601
[21] Gardener W A, Spooner C W. Cyclic spectral analysis for signal detection and modulation recognition. MILCOM'98, Vol.2: 419-424
[22] 韩国栋 蔡斌 邬江兴.调制分析与识别的谱相关方法. 系统工程与电子技术。2001,23(3):34-36
[23] 曹志刚.钱亚生.现代通信原理.北京:清华大学出版社,1992
[24] 王生兵.无线电信号的调制识别研究:[硕士学位论文].东南大学,2006
[25] 飞思科技产品研发中心,小波分析理论与 MATLAB7 实现,北京:电子工业出版社,2005.
[26] 张晓文,杨煜普,许晓鸣.基于小波变换的特征构造与选择.计算机工程与应用.2003,19: 25-28
[27] Learned R E, Willsky A S. A wavelet packet approach to transient signal classification[J]. Applied and Computational Harmaric Analysis, 1995: 2(3): 256-278
[28] Mallat S, Zhong S. Characterization of signals from multiscale edges[J]. IEEE Trans Pattern Recognition and Machine Intelligence,1992: 14(7):710-732
[29] Senhadji L, Carrault G, Ballanger J J, et al. Comparing wavelet transforms for recognizing cardiac patterns[J].IEEE Engineering in Mechine and Biology,1995:167-172
[30] Hazarika N, Chen J Z, Tsoi A C. Classification of EEG Signals Using the Wavelet Transform. IEEE DSP,1997:89-92
[31] Pittner S, Kamarthi S V. Feature From Wavelet Coefficients for Pattern recognition Tasks[J].IEEE Trans on Pattern Analysis and Machine Intalligence,1999:21(1):83-88
[32] Rioul O, Vetterli M. Wavelets and Signal Processing[J].ASSP Magazine,1991:14-38
[33] Shensa M J. The Discrete Wavelet Transform: Wedding the A-Trous and Mallat Algorithms[J]. IEEE Trans on Signal Processing,1992;40(10):2464-2478
[34] 胡广书.现代信号处理教程. 北京:清华大学出版社,2004
[35] 张贤达.现代信号处理.清华大学出版社(第二版),2002:378-428
[36] Avci, E., Turkoglu, I., Poyraz, M. (2005a). Intelligent target recognition based on wavelet packet neural network. Experts Systems with Applications, 29(1)
[37] Avci, E., Turkoglu, I., Poyraz, M. (2005b). In Intelligent target recognition based on wavelet adaptive network based fuzzy inference system. Lecture notes in computer science. Berlin: Springer-Verlag, May. 2005(3522):594–601
[38] Avci E. An expert Discrete Wavelet Adaptive Network Based Fuzzy Inference System for digital modulation recognition. Expert Systems with Applications.2007, 33(3),582-589
[39] Avci E. Performance comparison of wavelet families for analog modulation classification using expert discrete wavelet neural network system .Expert Systems with Applications. 2007, 33(1),23-35
[40] 夏克文,智能信息处理.天津:河北工业大学,2004
[41] 焦李成,神经网络系统理论[M],西安:电子科技大学出版社,1992
[42] 靳蕃 编著,神经计算智能基础,成都:西南交通大学出版社,2000 年
[43] 徐秉铮,张百灵,韦岗.神经网络理论与应用.华南理工大学出版社,1994
[44] Powell M J D. Radial basis functions for multivariate interpolation. In: Mason J M, Cox M, eds. Algorithms for Approximation.1985, 143-167
[45] Broomhead D S and D Lowe. Multivariate functional interpolation and adaptive networks. Complex Systems,1988(2):321-355
[46] 夏克文,宋建平.应用带有非线性连接权的神经网络识别水泥胶结质量.西安交通大学学报. 2003,37(2):192-195
[47] 飞思科技产品研发中心,神经网络理论与 MATLAB7 实现.北京:电子工业出版社,2005.
[48] 罗利春.无线电侦察信号分析与处理.国防工业出版社,2003:81-85
[49] Blake R.现代通信系统.电子工业出版社,2003:371-394
[50] Wong M L D, Nandi A K. Automatic digital modulation recognition using artificial neural network and genetic algorithm. Signal Processing, 2004,84:351–365
[51] Wu Z, Ren G, Wang X, et al. Automatic digital modulation recognition using wavelet transform and neural networks. ISNN 2004, LNCS 3173 :936–940
[52] Wu Z, Wang X, Liu C, et al. Automatic digital modulation recognition based on ART2A-DWNN. ISNN 2005, LNCS 3497 :381–386
[53] Zhang Q, Benveniste A. Wavelet network .IEEE Transactions on Neural Networks. 3(6):889–898
[54] 李建新,刘乃安,刘继平.现代通信系统分析与仿真-MATLAB通信工具箱.西安:西安电子科技大学出版社,2000.154-168
[55] Zhang J, Walter G G, Miao Y, et al. Wavelet neural networks for function learning. IEEE Transactions on Signal Processing, 43(6)
[56] Musavi M T, Ahmed W, Chan K H, et al. On the training of radial basis function classifiers, Neural Networks,1992. 595-603