Repeated joint source-channel decoding with reduced new bit errors

In this study, we describe a repeated joint source-channel decoding scheme in which not only a priori information about the bit to be decoded but also its decoding reliability information are used to adaptively control the decoding process. The channel decoding reliability information obtained for instance from a previous decoding attempt is used to properly weight the a priori information so that the new bit errors caused by wrong “natural” a priori information are reduced. In order to efficiently estimate the a priori residual redundancy, an iterative approach based on the Kalman filter theory is employed. As a result, decoding gains can be obtained with almost no new bit errors introduced. The proposed scheme is successfully applied to the GSM full rate system to achieve a more reliable transmission of speech signals in both noisy and less noisy environments.