基於類神經的波束成型技術應用於4x4正交分頻多天線系統的深衰落減免

Abstract

數位波束成型技術可以有效地對抗干擾訊號和多重路徑效果。利用波束成型技術生成權重來調整天線陣列可以有效地運用天線多重性來調整接收訊號的振幅和相位，以針對接受訊號到達接收端實的到達角度。 本論文針對多輸入多輸出的傳輸系統提出基於類神經網路的適應性數位波形技術架構，用意在於快速消除多重路徑效應產生的深衰弱現象。本論文提出的演算法事基於能量關聯性和類神經網路來計算數位波束成形的權重。利用能量關聯性先選取出適合的波束成形權重，再利用類神經網路在有限的循環次數內收斂出可消除深衰弱的波束成形權重。 在基於SCM和數位波型傳輸通道中模擬，基於4*4 MIMO OFDM的Wireless Backhual平台可以顯示出此演算法可有效地消除通道中所產生的深衰弱效應。在SNR=20且類神經網路的循環次數限制為10之下，可以有效地解決63%的深衰弱效應。

Digital beamforming is known to have interference rejection and capability against multipath effect when applying the precise steering array vector to antenna array. Steering array vector is carried out by weighting received digital signals, thereby adjusting their amplitudes and phases to form the desired beam toward AOAs (Angles of Arrival) of desired signals. In this paper, we propose a neuron-based robust adaptive beamforming in MIMO-OFDM system to solve the deep fading effect. The propose algorithm is based on correlation of power with table look-up and neural network which can accurately iterated to calculate beamforming weighting pattern. Use power correlation algorithm to choose an initial weighting patterns, and then use neural network to converge weightings pattern closed to steering vectors in order to decrease deep fading effect of multipath channel. By simulation in MIMO 4-by4 OFDM wireless backhaul system indicates that the proposed algorithm can solve deep almost 63% fading effect under SNR=20 and iteration limit=10 for neural network.