Near-field source arrival angle estimation method based on neural network

Abstract

A near-field source arrival angle estimation method based on a neural network comprises the steps that sampling data generated by K signals in a training interval are received by an array to calculatethe receiving data of the array, the signals are integrally rotated to acquire a next group of receiving data until the training interval is traversed; the covariance matrix of all groups of receiving data is established; a training data set is established through the covariance matrix and the arrival angles of the signals; the neural network is trained through the training data set; testing dataare calculated by sampling data generated by testing signals in testing angle intervals in the training interval are received by a sensor array, after one group of testing data are acquired, the signals are rotated to acquire the next group of testing data, the covariance matrix of all the groups of testing data is established, and normalization is conducted to obtain a normalized data covariancematrix; and the data of the normalized data covariance matrix are input into the neural network to calculate the arrival angle estimated value of the testing signals.

Description

technical field [0001] The invention belongs to the technical field of signal processing, in particular to a method for estimating a one-dimensional arrival angle of a near-field source. Background technique [0002] Arrival angle estimation is the core of space source passive direction finding technology and an important research topic in array signal processing. The traditional angle of arrival estimation is mainly carried out by MUSIC method, ESPRIT method and its derivative methods. The MUSIC method decomposes the array received data into a signal subspace and a noise subspace, and then uses the steering vector to scan the noise subspace to estimate the angle of arrival. Since the steering vector must scan the entire space during this process, the amount of calculation is huge, which makes it impossible to realize real-time processing of data. Moreover, the MUSIC method has poor performance in estimating the angle of arrival in the case of low signal-to-noise ratio and...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the problems that MUSIC method and ESPRIT method exist in the non-coherent signal with low signal-to-noise ratio, the angle-of-arrival estimation effect of the small interval angle is not good and the coherent signal is completely invalid. The purpose of the present invention is to provide a near-field source angle-of-arrival based on neural network Estimation method, which can improve the angle of arrival estimation effect of coherent signals and non-coherent signals under low signal-to-noise ratio

Images