Rao-Blackwellized particle forward filtering backward smoothing with application to blind source separation
Hao Qiu, Gaoming Huang, Jun Gao · 2014
It is difficult for standard independent component analysis (ICA) algorithm to extract signals in noise condition. The main contribution of this paper is to apply Rao-Blackwellized particle filter and smoother to noisy ICA model. The proposed method is presented as a two-stage approach. Firstly, noisy signal is modeled by time-varying autoregressive (TVAR) process, and estimated noise-free signal is obtained by particle filtering and smoothing step. After the preprocessing step, Fast ICA algorithm is adopted to separate the denoising data. The enhancement performance of proposed algorithm is evaluated in simulations at last.