小波法結合類神經網路之雷達信號源識別

Identification of Radar Transmitter Using Wavelet Transform and Neural Network

葉金益(Chin-Yi Yeh)

指導教授 ： 陳永盛

元智大學/工程學院/電機工程學系/碩士(2004年)

摘要

本研究是利用小波法具備分析暫態信號強大的功能與效率，對信號源的發射訊號進行分析與特徵萃取。由於（i）小波轉換具有局部（local）處理信號的能力，對於瞬間變化的暫態信號能有效的凸顯其特徵。（ii）小波具有頻率特性，使得在處理多樣多變的複雜信號，也能輕易的掌握其特性。（iii）相對於頻率域之轉換方法，小波轉換處理速度快（因不須事先經過繁雜的訓練與數學計算）又能在每一層級近似的轉換下，將資料量遞減為1/2n，使得小波轉換在速度處理上有較佳的效率。基於上述優點，本研究採用Coif.小波族群轉換技術，將偵搜之訊號萃取特徵值，最後以倒傳遞類神經網路辨識法，快速、高效率的辨識信號源脈波序列信號。並以數百種信號進行測試、驗證、評估，確認辨識效果與執行速度均十分良好。

關鍵字

小波轉換

並列摘要

This research utilizes wavelet method’s powerful transient signal function and efficiency to carry out analysis and characteristics extraction of signal source’s signal transmission. (i) Since wavelet conversion has local signal processing ability, the characteristics of the momentarily changing transient signals can be highlighted. (ii) Since wavelet has frequency characteristics, it can easily keep track of the characteristics while processing varsity and changing diversified and changing complex signals. And, (iii) comparing against frequency domain conversion method, the speed of wavelet conversion processing is quicker (it does not need complex training and mathematical calculation in advance), under approximate conversion at each level, it reduce information amount by 1/2n, the wavelet conversion has better efficiency in speed processing. Based on the above-described advantages, this research utilizes the Coif. wavelet group conversion technology to extract the characteristic values of the signals. It then uses reverse transmission neural network recognition method to recognize signal source’s pulsation serial signals with high speed and high efficiency. It utilized hundreds signals for testing, identification and evaluation. It verified that recognition results and execution speed are excellent.

並列關鍵字

wavelet transform

參考文獻

[1] M. P. Fargues, W.A. Brooks, "Comparative study of time-frequency and time-scale transforms for ultra-wideband radar transient signal detection" IEE Proc.-Radar, Sonar Navig., Vol.142, No. 5, pp.236-242 October 1995.

[2] D. Sersic, "Wavelet filter banks with adaptive number of zero moments "Signal Processing Proceedings, WCCC-ICSP2000.5th International Conference vo1.1 pp.325-328 Aug.2000.

[4] S. G. Mallat, "A theory for multiresolution signal decomposition: the wavelet representation", IEEE Transactions on pattern & machine intelligence, Vol. 11,No. 7. pp.674-693 July 1989.

[5] R. W. Ehrich, J. P. Foith, "Representation of random waveforms by relational trees" IEEE Trans. on computers Vol. C-25, No.7 pp.725-736 July 1976.

[6] R. J. Mitchell, R.C. Gonzalez, "Multilevel crossing rates for automated signal classification" Proceding IEEE ICASSP, pp.218-222, 1978.