The Maximum A Posteriori Decoding Using Variational Bayes Methods for Digital Magnetic Recording Channels

Abstract

In this paper, we present new maximum a posteriori (MAP) decoding based on the Variational Base (VB) method in digital magnetic recording channels with colored Gaussian noise and transition noise. Multivariate autoregressive (AR) models is used for an inference model for the mixture noise sources. Noise estimating in the MAP detector uses the VB method of the AR mixtures. It is shown that the proposed MAP detector is effective to estimate signal sequences degraded by the mixture noise sources and improve the bit error rate (BER) performances.