利用多重演算法於多輸入多輸出正交分頻多工系統通道估測之研究

Research on Channel Estimation in MIMO-OFDM System via A Multi-Algorithm

黃信偉(Hsin-Wen Huang)

指導教授 ： 鄭佳炘

國立虎尾科技大學/電機資訊學院/電機工程研究所/碩士(2015年)

https://doi.org/10.6827/NFU.2015.00026

摘要

正交分頻多工(Orthogonal Frequency Division Multiplexing，OFDM)是一種具有高速資料傳輸能力與良好的頻寬使用效率(bandwidth efficiency)的無線通訊技術。而無線通訊傳輸一定會遇到的問題就是訊號經過無線通道後會因為通道效應或雜訊等干擾使得訊號有衰減、累加、扭曲與變形…等現象，接收端收到的訊號會與原本所傳送的訊號不同，系統效能就會變差，為了使系統效能變好，必須再接收端做通道估測來將訊號補償，才能收到傳送端所傳送的正確訊號。本論文在多輸入多輸出正交分頻多工(Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing, MIMO-OFDM)系統下研究利用不同演算法來補償訊號並比較其效能。
人工神經網絡(artificial neural network, ANN)現階段已廣用於各個領域了，倒傳遞類神經網路(Back Propagation Neural Network, BPNN)為人工神經網絡的一種，用於通道估測可以幫助我們處理因雜訊干擾而無法完美處理線性計算的難題。但本論文使用倒傳遞類神經網路發現在各種條件下，其位元錯誤率都很高，主要原因是因為多輸入多輸出系統是利用空-時塊編碼(Space-time block code, STBC)所架構而成，所以接收端收到訊號時是兩個傳送訊號相加而成，對於倒傳遞類神經網路來說是無法將其分解再個別訓練資料的，因此單純用倒傳遞類神經網路所估出來的效果很差。本論文提出將最小平方法(Least-Square, LS)與倒傳遞類神經網路兩者結合，就可以解決因兩訊號相加而造成無法有效利用倒傳遞類神經網路來訓練資料的問題。
本論文是在瑞利衰減通道(Rayleigh fading channel)通道環境下，分別觀察領航符元資料(Pilot)的數量及不同學習率(Learning rate)下的位元錯誤率。最後本文會利用最小平方法結合倒傳遞類神經網路(LS-BPNN)與最小平方法、最小均方誤差法(Minimum Mean-Square Error, MMSE)和倒傳遞類神經網路進行通道估測來補償資料並比較它們的系統效能。模擬結果顯示，將兩種演算法合併使用確實可以達到較好的系統效能，也能解決倒傳遞類神經網路無法訓練復合訊號的問題。

關鍵字

正交分頻多工 ； 通道估測 ； 多輸入多輸出 ； 倒傳遞類神經網路 ； 空-時塊編碼 ； 最小平方法

並列摘要

Orthogonal Frequency Division Multiplexing (OFDM) is a high data rate wireless communication technique. The wireless communication must suffer a problem that the signal will be distorted after passing through the wireless channel. The received signals are different from the transmitted ones, lead to poor system efficiency. To enhance it, the channel estimation in the receiver is necessary. The thesis researches on channel estimation in Multi-Input Multi-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) system via a multi-algorithm.
Artificial Neural Networks (ANNs) are popular in each territory. Back Propagation Neural Network (BPNN) is one of them. BPNN is able to handle the nonlinear computation. The thesis utilizes BPNN for channel estimation; however, the bit error rate is very poor. Since our MIMO system is constructed from Space-Time Block Code (STBC), the received signals would be a complex signal. BPNN can't take them apart then trains data respectively. Thus, BPNN has a poor result for channel estimation in this system.
The thesis proposed the combination of Least-Square (LS) and BPNN. The approach solves the problem that BPNN isn't capable of training data effectively. The thesis observes the bit error rate in pilot symbols of various length number of and learning rate on Rayleigh fading channels. Eventually, we compensate the bit error rates and mean square error with Least-Square combined with Back Propagation Neural Network (LS-BPNN), LS, Minimum Mean-Square Error (MMSE) algorithms. The simulation results express that combining two algorithms has a better performance, and has a solution for BPNN about training multiple signals.

並列關鍵字

OFDM ； channel estimation ； MIMO ； BPNN ； STBC ； LS

參考文獻

[14] 鄭永沛, “在OFDM系統以基於基因演算法之倒傳遞類神經網路進行通道估測之研究,”國立虎尾科技大學電機工程研究所碩士論文, pp. 1-80, 2012.

[13] 李文慶, “在OFDM系統中以改良式倒傳遞類神經網路進行通道估測之研究,” 國立虎尾科技大學電機工程研究所碩士論文, pp. 1-126, 2014.

[23] P. K. Pradhan, S. K. Patra, O. Faust, and B. K. Chua, “Channel estimation algorithms for OFDM systems,” International Journal of Signal and Imaging Systems Engineering, vol. 5, no. 4, pp. 267-273, 2012.

[25] C. H. Cheng, Y. P. Cheng, Y. H. Huang, and W. C. Li, “Using Back Propagation Neural Network for Channel Estimation and Compensation in OFDM Systems,” in Proc. of 2013 Seventh International Conference on Complex, Intelligent, and Software Intensive Systems (CISIS), pp. 340-345, Taichung, Taiwan, Jul. 3-5, 2013.

[1] L. J. Cimini, “Analysis and simulation of digital mobile channel using orthogonal frequency division multiplexing,” IEEE Transactions on Communications, vol. 33, no. 7, pp. 665-675, Jul. 1985.