Frequency-selective and nonlinear channel estimation with unknown noise statistics

We propose cost reference particle filter (CRPF) and extended game theory-based H_¿ filter approaches to the problem of estimating frequency-selective and slowly varying nonlinear channels with unknown noise statistics. The proposed approaches have a common advantageous feature that the noise information is not required in their applications. The simulation results justify that both approaches are effective, and that CRPF is more robust against highly nonlinear and drastically varying channels.