一种基于傅里叶基神经网络的频谱分析方法

曾喆昭

doi:10.3724/SP.J.1146.2008.01118

一种基于傅里叶基神经网络的频谱分析方法

doi: 10.3724/SP.J.1146.2008.01118

曾喆昭

基金项目:

湖南省教育厅重点课题(08A006)资助项目

计量
- 文章访问数: 3392
- HTML全文浏览量: 110
- PDF下载量: 836
- 被引次数: 0
出版历程
- 收稿日期: 2008-09-09
- 修回日期: 2009-03-16
- 刊出日期: 2009-08-19

A Spectrum Analysis Method Based on FBF Neural Network

Zeng Zhe-zhao

摘要

摘要: 该文提出了一种用递推最小二乘法训练傅里叶基神经网络权值的频谱分析方法。其主要思想是采用递推最小二乘法训练傅里叶基神经网络权值，根据权值获得信号的幅度谱和相位谱。该方法不涉及复数的乘法运算和加法运算，便于软件和硬件实现，特别适合于DSP软硬件实现。仿真结果表明，该方法不仅计算精度高，计算速度快，而且具有噪声滤波功能，是一种有效的频谱分析方法。
- 频谱分析;神经网络;递推最小二乘法
Abstract: A method of spectrum analysis is proposed using recursive least square algorithm to train the weights of Fourier Basis Functions (FBF) neural network. Its main idea is that the weights of the FBF neural network are trained with recursive least square algorithm, and then the amplitude spectrum and phase spectrum are computed according to the obtained weights. The approach is not involved in plural multiplication operation and plural addition operation. Software and hardware of the approach are easily implemented. It is especially fit for the implementation of software and hardware of DSP chips. The results of spectrum analysis show that the method presented has not only high accuracy and fast convergence speed but also filtering function. Therefore, it is an effective approach of spectrum analysis.

HTML全文

参考文献(1)

He Yan-feng, Wu Xiao-dong, and Han Guo-qing, et al..Frequency spectrum analysis method for recognition ofdynamometer card. Acta Petrolei Sinica, 2008, 29(4):619-624.[2]Lee Usik, Lee Jaesang, and Kim Chang-Boo. Spectralanalysis method for nonlinear dynamic systems. 10thInternational Conference on the Mechanical Behaviour ofMaterial, Switzerland, 2007: 861-864.[3]Raghavan S, Imbrie P K, and William A. Spectral analysis ofR-lines and vibronic sidebands in the emission spectrum ofruby using genetic algorithms[J].Applied Spectroscopy.2008,62(7):759-765[4]Lin Wei-jun. Chebyshev spectral element method for elasticwave modeling. Acta Acustica, 2007, 32(6): 525-533.[5]侯正信, 王兆华, 杨希. 全相位DFT 数字滤波器的设计与实现[J]. 电子学报, 2003, 31(4): 539-543.Hou Zheng-xin, Wang Zhao-hua, and Yang Xi. Design andimplementation of all phase DFT digital filter[J]. ActaElectronica Sinica, 2003, 31(4): 539-543.[6]Yang Hai, Cheng Wei, and Chu Li-yan. Time-frequencyanalysis based on process neural network for non-stationaryrandom vibration of spacecraft. Journal of Vibration andShock, 2008, 27(1): 12-15.[7]申小芳, 姚骑均, 林镇辉等. 用外差混频和直接检波方法测量亚毫米波连续波源的频谱[J]. 红外与毫米波学报, 2005, 24(5):3-5.Shen X F, Yao Q J, and Lin Z H, et al.. Measurement of thefrequency spectrum of a sub-millimeter cw source byheterodyne mixing and direct detection[J]. Journal InfraredMillimeter and Waves, 2005, 24(5): 3-5.[8]梁唯一，焦扬，王均宏. 连续结构微带滤波器的矩阵运算法分析[J]. 红外与毫米波学报, 2006, 25(1): 62-65.Liang Wei-yi, Jiao Yang, and Wang Jun-hong. Analysis ofstructure-continuous micro-strip filters by matrixmultiplication method[J]. Journal Infrared Millimeter andWaves, 2006, 25(1): 62-65.[9]马艳华, 王建宇, 马德敏, 舒嵘. 一种用于光谱图像的基于邻域背景检测的矢量滤波器[J]. 红外与毫米波学报, 2006, 25(2):79-82.Ma Yan-hua, Wang Jian-yu, Ma De-min, and Shu Rong. Newvector filter based on background detection used in spectralimages [J]. Journal Infrared Millimeter and Waves, 2006,25(2): 79-82.[10]Katsuhiro I S and Katsuhiro I W, et al.. Numerical analysis ofa path-length-resolved spectrum of time-varying scatteredlight field[J].Journal of the Optical Society of America A:Optics and Image Science, and Vision.2008, 25(3):718-724[11]曾喆昭, 竺炜, 王耀南. 一种基于神经网络算法的频谱分析方法[J]. 红外与毫米波学报, 2007, 26(2): 141-145.Zeng Zhe-zhao, Zhu Wei, and Wang Yao-nan. Analysisapproach of spectrum based on neural network algorithm [J].Journal Infrared Millimeter and Waves, 2007, 26(2): 141-145.[12]曾喆昭, 文卉, 王耀南. 一种高精度的电力系统谐波智能分析方法[J]. 中国电机工程学报, 2006, 26(10): 23-27.Zeng Zhe-zhao, Wen Hui, and Wang Yao-nan. An approachwith high accuracy for intelligent analysis of power systemharmonics [J]. Proceedings of the CSEE, 2006, 26(10): 23-27.

施引文献

资源附件(0)

访问统计