Font Size: a A A

Design Of A Platform For Estimating The Number Of Array Signals Based On Machine Learning

Posted on:2021-03-28Degree:MasterType:Thesis
Country:ChinaCandidate:X L ZhangFull Text:PDF
GTID:2438330602997666Subject:Electronics and Communications Engineering
Abstract/Summary:PDF Full Text Request
Spatial spectrum estimation technology mainly estimates the number of signals and the direction of arrival(DOA).Among them,most super-resolution spectral estimation algorithms are based on the premise of correctly estimating the number of signals.If the number of signals is incorrectly estimated,the direction of arrival will be greatly affected.In the background of spatial colored noise,the paper studies the problem of performance degradation of the estimation algorithm of independent signals and coherent signals when the signal noise is relatively low and the number of snapshots is small.The paper introduces the basic signal model of uniform linear array(ULA),color noise model and coherent signal model,analyzes the influence of the number of signals on DOA under three conditions through simulation experiments,and then introduces several signal number estimation algorithms.Secondly,under the background of colored noise,when the received signal is an independent signal,the paper proposes an estimation algorithm using support vector machine(SVM)to estimate the number of signals.Use the difference of the signal and noise in the received data to extract features,train the SVM classification model,and then use the trained classification model to estimate the number of unknown signals.When the received data contains coherent signals,this method can be used to estimate the number of signals after decoherence.Compared with other algorithms,it proves that the proposed algorithm is effective.Then,when the received signal contains coherent signals,an algorithm for back propagation(BP)neural network to estimate the number of signals is proposed,and the particle swarm optimization(PSO)is used to solve the neural network training Problems of instability or poor convergence.The algorithm first extracts the relevant feature parameters,then uses the feature parameters as the input training model of the neural network,and finally estimates the number of signals in the test data,and this method can also estimate the number of coherent signals.Through experiment comparison with other algorithms,it can be proved that the proposed algorithm has good estimation performance.Finally,the digital signal processor(DSP)is used to realize two kinds of signal number estimation algorithms.The implementation of the algorithm is analyzed separately,and the real-time test results show that the signal number estimation algorithm in the paper can achieve good estimation results and has certain engineering practicability.
Keywords/Search Tags:Signal number estimation, support vector machine, BP neural network, DSP
PDF Full Text Request
Related items