SOURCE ENUMERATION ALGORITHM BASED ON EIGENVECTOR: REVISIT FROM THE PERSPECTIVE OF INFORMATION THEORY

Wenzhun Huang, Shanwen Zhang

Department of Engineering Technology, Xijing University, No.1 Xijing Road, Xi’an, 710123, China

 In case of low signal to noise ratio (SNR) and small snapshot condition, it is difficult to separate sources and noises, and the performance of classical eigenvector source estimation algorithm drops quickly. To solve the problem, further research is carried out around the characters of eigenvalue and eigenvector, and a novel eigenvalue algorithm is presented based on the theory of source enumeration. In detail, the eigenvectors of sample covariance matrix are employed as the decision factor, which is insensitive to SNR. And an improved Predictive Description Length (PDL) criterion is adopted to enumerate source number. Theoretical analysis and simulation results demonstrate that the proposed algorithm is available and efficient in case of low SNR and small snapshot condition compared with those of Minimum Description Length (MDL) and PDL.