In previous work on source coding over noisy channels, it was recognized that when the source is correlated, "residual redundancy" typically remains between the discrete symbols produced by the encoder. This inherent redundancy can be capitalized upon by the decoder to improve the overall quantizer performance. Sayood and Borken-hagen and Phamdo and Farvardin both proposed sequence-based "detectors" at the decoder which optimize suitable criteria in order to estimate the sequence of transmitted symbols. Phamdo and Farvardin also proposed an instantaneous, approximate minimum mean-squared error (MMSE) decoder. While these methods have been shown to provide a performance advantage over conventional systems, the former decoder structure (detector-based) is sub-optimal, while the latter structure makes limits use of residual redundancy in the sequence. Alternatively, combining both approaches, we propose a sequence-based, approximate MMSE decoder which utilizes the entire observation sequence and computes expected values based on a discrete hidden Markov model. Significant performance gains are demonstrated over previous techniques in quantizing Gauss-Markov sources, over a range of noisy channel conditions. Moreover, constrained versions of the new technique are suggested in order to limit the system delay.
展开▼