Marginalized Particle Filtering for Blind System Identification Journal Articles

This thesis develops a marginalized particle filtering algorithm for the blind system identification problem. The blind system identification problem arises in many fields, including speech processing, communications, biomedical signal processing, sonar and seismology. The state space model under consideration uses a time-varying autoregressive (AR) model for the sources, and a time-varying finite impulse response (FIR) model for the channel. The multi-sensor measurements result from the convolution of the sources with the channels in the presence of additive noise. A numerical approximation to the optimal Bayesian solution for the sequential state estimation problem is implemented using the particle filter. Estimates of the sources are recovered directly by marginalizing the AR and FIR coefficients out of the posterior distribution for the unknown system parameters. The resulting marginalized particle filtering algorithm allows efficient identification of the system. Simulation results are given to verify the performance of the proposed method. The block sequential importance sampling (BSIS) formulation of the particle filter is also introduced to exploit the structure inherent in the convolution state space model.