Mixing matrix estimation algorithm based on single source detection
Donghui Xu, Yang Dongdong, Haomiao Liu, Hongguang Ma · 2016
In this paper, the input signal's short time Fourier transformation and the sparsity condition satisfied in the time-frequency domain are given firstly. Then, the feasibility of the single source detection using the wideband radar signal based on the eigen-value decomposition is analyzed theoretically. Finally, the system clustering method is used to estimate the number of sources and the mixing matrix by analyzing the eigen-vector corresponding to the single source neighborhood using the clustering method. For the reason that the clustering analysis is done by using the eigen-vectors instead of the observed signals, the estimation of the mixing matrix using this method has a high precision and strong robustness, and it overcomes the defects of the traditional K-means clustering algorithm that the analyzed signal must be sufficiently sparse and the number of the source signal must be known in advance, etc.