A new algorithm for cyclostationary signal blind deconvolution

Blind source separation is to retrieve sources from mixed signals collected by sensors, without any prior knowledge about the sources and the pipeline. Some source signals is cyclostationary signal. The proposed method is derived by applying nonholomic constraints natural gradient iterative learning to the novel contrast function which is defined according to convolutive mixtures and cyclostationary of source signals. The efficiency of the algorithm is supported by simulation, which shows that the proposed method has improved performance for the separation in terms of waveform similarity, separation filter global transfer filter and convergence speed, compared to the traditional deconvolution method.