Algorithm for Multichannel Blind Deconvolution Based on Subspace Decomposition

付卫红,杨小牛,刘乃安,曾兴雯

DOI: https://doi.org/10.3321/j.issn:1000-436x.2009.01.005

2009-01-01

Abstract:Aim at the blind separation of convolution mixture signals, a novel algorithm for multi-channel blind deconvolution was presented. Firstly, this algorithm transforms the convolution mixture into the instantaneous linear mixture by subspace decomposition. The second step is to separate the estimated linearized signals and to obtain the source signals by using the blind source separation algorithm for instantaneous linear mixture. This algorithm needs not to solve the problem of the scale and permutation ambiguity existing in the instantaneous linear blind separation relative to the frequency domain method for blind deconvolution. Moreover, the algorithm doesn't need signal is independent identically distributed(i.i.d), only needs the sources are statiscally independent each other. Therefore the algorithm can process the mixing signal in intermediate frequency. Simulation results show that this algorithm can blind deconvolute not only the signals with the different carrier frequencies and different symbol rates, but also that with same carrier frequencies and same symbol rates.