多層感知決策回授等化器之調適性活化函數分析

數位通訊系統中，訊號藉由發射端經由通道傳遞至接收端時，由於自然界中的通訊通道大多非理想性。因此，通道會造成失真現象的產生。一般而言影響失真現象的因素，較常見的有：符碼間造成的符際干擾(Inter-Symbol Interference, ISI)、共通道干擾(Co-Channel Interference, CCI)、通道中的雜訊(Noise)和多重路徑存取干擾(Multi-path Interference, MAI)等。傳統上降低干擾因素的作法，是在通道接收端設計等化器(Equalizer)加以改善。為降低通道內符際、雜訊等干擾，使通道特性更趨穩定，使用調適性濾波器(Adaptive Filters)的想法便產生。調適性濾波器可依據通道環境變化，自我調整並適應環境變化。依不同等化器架構搭配演算法可得到不同程度的改善。使用類神經網路(Neural Networks)架構的多層感知決策回授等化器(Multilayer Perceptron Decision Feedback Equalizer, MLP-DEF)，可得到較佳的改善情況。因多層感知決策回授等化器輸出層節點，大多使用固定形式之活化函數(Activation Function)，雖已有不錯的性能表現。但本論文針對輸出層節點的活化函數提出另一種思維，企圖調整參數值使系統收斂速度加快，並探討誤碼率(Bit Error Rate, BER)的變化情況。本論文於不同訊號雜訊比(Signal Noise Ratio, SNR)條件下，經程式模擬實驗後加以討論、比較調整活化函數(Activation Function)前、後，對系統的收斂特性、誤碼率所產生的影響。

關鍵字

活化函數；等化器；調適性濾波器；多層感知決策回授； Sigmoid函數

並列摘要

In digital communication system, digital signal will distort while the signal is sent to the receiver through the channels because all the communication channels in nature are not ideal. Generally, reasons causing distortion include Inter-Symbol Interference (ISI), Co-Channel Interference (CCI), Noise, Multi-path Interference (MAI), etc. Traditionally, to lessen interferences, an equalizer is installed on the receiver to improve the distortion. On the other hand, to reduce ISI and Noise, and make channel features more stable, Adaptive Filters is applied as it will self-adjust to fit the situation of channel keeping changing all the time and with algorithm, different equalizer structure results in different improvement. Multilayer perceptron decision feedback equalizer (MLP-DEF) of Neural Networks structure makes the interferences better improved. Although most of the MLP-DEF output layers, using fixed activation function, performs well, this article is trying to address a different idea in the light of this activation function to accelerate system convergence rate by adjusting the parameter, , and to explore how Bit Error Rate (BER) changes. Furthermore, we want to know the effect on this convergence rate and BER under the condition of different SNR (Signal Noise Ratio), by way of discussing the simulation experiments of formula, and comparison between before and after the adjustment.

並列關鍵字

activation function ； equalizer ； adaptive filter ； multilayer perceptron decision feedback ； sigmoid function

參考文獻

[20]楊盛松，”複數訊號多層感知回授等化器-使用進化演算法”，國立中央大學電機工程研究所碩士論文，Chapter 5, 2002.

[1]Qureshi, S.U.H, “Adaptive Equalization”, Proceeding of the IEEE, vol.73, No.9, pp.1349-1387, Sep. 1985.

[2]G.J. Gibson, S. Siu, and C.F.N Cowan, “The application of nonlinear structures to the reconstruction of binary signals”, IEEE Transactions on Signal Porcessing, vol.39, No.8, pp.1877-1884, Feb. 1991.

[3]P. Sarwal & M.D. Srinath, “A fuzzy Logic System for Channel Equalizer”,IEEE Transactions on Fuzzy Systems, vol.3, No.2 , pp.246-249, May 1995.

[7]W. McCulloch and W. Pitts, “A logical calculus of the ideas immanent in nervous activity”, Bull. Math. Biophysics, 5 (1943), 115-133.

國際替代計量

多層感知決策回授等化器之調適性活化函數分析

全文下載

主題瀏覽