An Improved DOA Estimation Algorithm of Neural Network Based on Interval Division

As neural network algorithm springs up, it is being applied to the DOA estimation more and more often. However, the robustness and real-time performance of the neural network algorithm under the condition of the multiple sources has always been a difficult problem. The DOA estimation algorithm of the neural network based on interval division has better robustness and real-time performance, which divides the signal into different angle intervals. However, the accuracy of signal division is difficult to be guaranteed, especially when the angle of signal is at interval edge, causing the large error. An improved interval division method is thus proposed in this paper. Firstly, by adjusting the weight of the antenna array, the beam of the antenna array is focused at the angle corresponding to each sub-region, therefore the spatial signal feature is improved, In the interval division, the edge overlapping division is adopted to improve the estimation accuracy of the interval edge angle. After experimental verification, the improved interval partition method can improve the performance of the algorithm.