Bayesian blind deconvolution for mobile communications
T.C. Clapp · 1997
This paper discusses the problem of restoring a digital input signal which has been degraded by an unknown FIR filter in additive Gaussian noise. A Bayesian approach is taken to recover the signal, implemented by the Gibbs sampler, a Markov chain Monte Carlo method. A method for drawing a random sample of a sequence of bits is presented. This is shown to have a faster convergence and better performance than a scheme by Chen and Li (see IEEE Transactions on Signal Processing, vol.43, no.10, p.2410-13, 1995) which draws bits independently. (6 pages)