| Array signal processing,as an important branch of signal processing,has extensive applications in radar,communications,sonar and biomedical engineering.One of the basic problems is the direction of arrival,DOA.The signal environment is becoming increasingly complex.The signal density is increasing,and the frequency range of the signal is continuously expanding.As a result,the distribution range and density of the signal in the airspace and frequency domain are greatly increased,and the shortcomings of the narrowband system are gradually displayed.The echo of the wideband signal carries a large amount of information,so wideband technology is increasingly used,but the realization of the broadband technology is more difficult than the narrowband.Coupled tensor decomposition,as a new multi-set high-dimensional data fusion tool,has gradually gained attention due to its great potential in identification and resolution accuracy,if it can be combined with the direction finding of wideband array signals,it will greatly promote the development in theory,methods and applications.The article will solve the DOA estimation problem of wideband array signals from the perspective of coupled tensor decomposition.For a typical small array(spatially stretched quint dipole,SS-Quint)and large array(large-scale uniform linear array)structures,tensors are constructed using either statistical domain or data domain tensorization methods,and DOA estimation is performed by using coupled canonical polyadic decomposition algorithms.The main content of this article is summarized as follows:For a small array: SS-Quint,we propose one wideband DOA estimation algorithm based on double coupled canonical polyadic decomposition.First of all,we perfected the vector cross product based on SS-Quint,expand the quadrant of the incident signal from the original quarter sphere to the full quadrant,solve the problem of symbol ambiguity,and extend it to the large arrays through an optimization algorithm.For large element spacing,the algorithm is still applicable.Further,we propose double coupled canonical polyadic decomposition for direction finding of wideband array signals.First,we transform the time domain accepted data into the frequency domain through Fourier transform,these bands can be viewed as multiple data sets,and tensorization is performed by calculating the matrix covariance in the statistical domain.Finally,we use double coupled canonical polyadic decomposition and vector cross product to perform DOA estimation.For large-scale uniform linear arrays,a DOA estimation algorithm for wideband array signals based on coupled canonical polyadic decomposition is proposed: First,we perform the Short-Time Fourier Transform(STFT)on the mixing matrix of the wideband array,and convert it to multiple narrowband data in the frequency domain,where each frequency band can be seen as a data set.Tensorization is performed by spatial sampling,and then we can perform the DOA estimation by using coupled canonical polyadic decomposition. |
