Direction Of Arrival Estimation Based On Signal Receiving Intensity Gradient

Passive direction-finding technology is widely used in various fields of national defense and national economy,and is one of the researches focus in signal processing.The classical passive direction-finding technology uses phase interferometer and super-resolution algorithm to realize DOA estimation by processing the amplitude and phase of the received signal from the array antenna.The performance of this kind of algorithm is limited by many factors,such as the incident signal form,the signal propagation environment,the arrangement of antenna array,the consistency of system amplitude and phase,etc.Aiming at these limitations of classical passive direction-finding technology,a passive direction-finding algorithm based on the received signal strength value under the condition of blind array antenna structure and signal form is used in this paper.The main content of the article is as follows:Firstly,in an ideal situation,this paper establishes the basic data set of the received signal strength at the randomly distributed receiver,and establishes the linear relationship between the RSS and the spatial coordinate according to the distribution law of the signal propagation in the two-dimensional space.The DOA estimation is transformed into the parameter estimation in the linear regression equation through the derivation of the free space distance loss model formula,and the RSS gradient of the receiver is estimated.According to the characteristics of large sample size and spatial attenuation distribution,this paper use Bayesian linear regression method to estimate the parameters.Secondly,aiming at the spatial correlation in RSS data set,Moran's I parameter is selected as the measurement index of data relevance,and the weight matrix of Moran's I is calculated based on the distance factor,combined with the improved K-means clustering algorithm which optimize the initial center of clustering,is used to detect and reduce the spatial correlation of the data set,to reduce the error data generated by the local attenuation.On this basis,the paper further analyzes the data singularity caused by the non-ideal factors such as occlusion,scattering or measurement error in the actual signal propagation situation,and tests the outliers both in space property and model property respectively.For the spatial outliers,the nearest neighbor rule based on distance is applied,and the data is filtered according to the difference of non-spatial attribute values of data points.For the model outliers,the improved RANSAC algorithm based on the optimized sample subset is applied to test the parameters,obtain the optimal model,and remove the outliers that do not conform to the estimation model.Finally,six groups of measured data in 2.4GHz and 5GHz frequency bands are used for simulation experiment.The experimental results show that the RSS distribution of the data set processed by the singular value of data is smoother and more suitable for the ideal distribution of signal propagation.Combined with the linear regression RSS gradient parameter estimation method,the direction of arrival can be estimated effectively.