Method and device for direction of arrival estimation based on cascaded neural network

Abstract

The embodiment of the present invention provides a method and device for estimating the direction of arrival based on a cascaded neural network. Based on the received signals from multiple signal sources, the covariance matrix of the received signal is determined; the eigenvalue decomposition is performed on the covariance matrix , to obtain the eigenvalue vector; convert the elements in the covariance matrix into normalized real vectors to obtain the covariance information vector; input the eigenvalue vector and covariance information vector into the pre-trained cascaded neural network to obtain The direction of arrival of the source signal; the cascaded neural network includes a signal-to-noise ratio classification network and a direction-of-arrival estimation network, and the direction-of-arrival estimation network includes a high-signal-to-noise ratio estimation sub-network and a low-signal-to-noise ratio estimation sub-network; The output result of the ratio classification network is high SNR activating the high SNR estimation subnetwork; the output result of SNR classification network is low SNR activating the low SNR estimation subnetwork. Can be applied to a wide range of signal-to-noise ratios.

Description

technical field [0001] The present invention relates to the technical field of wireless communication, in particular to a method and device for estimating a direction of arrival based on a cascaded neural network. Background technique [0002] DOA (Direction of Arrival Estimation, direction of arrival) estimation is an important research topic in the field of wireless communication, the purpose is to obtain the incident direction of the signal when it reaches the antenna. [0003] In the existing neural network-based DOA estimation method, the covariance matrix is ​​constructed based on the received signal, which contains all the signal DOA information, and the elements in the upper triangular matrix of the received signal covariance matrix are input to pre-construct and train In the completed neural network, the DOA estimation result of the output signal is the direction angle of the incident signal. However, this DOA estimation method has certain defects, which are manife...

AI Technical Summary

Problems solved by technology

However, this DOA estimation method has certain defects, which are manifested in the following two points: 1) It is not suitable for scenes with a large range of signal-to-noise ratios. If the noise-free signal data is used for network training, the network will The estimation performance is poor; if using noisy signal data for network training, although the network can obtain better estimation performance in the case of low signal-to-noise ratio, the estimation performance in the case of high signal-to-noise ratio is not as good as traditional DOA estimation algorithm

That is, poor adaptability to noise

2) DOA estimation can only be performed on signals sent by a single source, and cannot be applied to scenarios with multiple sources

Images