DOA Estimation with Artificial Neural Network

• Main Authors: Chen, You-Yu, 陳宥諭
• Other Authors: Wu, Wen-Rong
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/2pxvzy

碩士 === 國立交通大學 === 電信工程研究所 === 107 === A uniform linear array can receive multiple unequal power signals coming from different direction-of-arrival (DOA). This thesis considers DOA estimation with artificial neural network (NN). Conventional NN approaches are not effective for the unequal-power scenario. Also, the computational complexity is very high. Incorporating signal processing techniques, we propose two NNs to solve the problems. The first NN divides the estimation range into sectors, and consists of a spatial filter and a classifier. With a rotation operation, a spatial filter and a classifier can be used for all sectors, significantly reducing the training time and computational complexity. The second NN uses the same sector-based processing structure. However, the spatial filter is replaced with a power detector. With a frequency-domain nulling operation, only a power detector and a classifier are needed for all sectors. The computational complexity of the second NN can be further reduced. Simulation results show that the performance of the proposed NNs can outperform the well-known MUSIC algorithm under low SNR. Also, the computational complexity can also be lower than that of MUSIC.