Parameter SNR Optimized Index Assignments and Quantizers based on First Order A Priori Knowledge for Iterative Source-Channel Decoding

Two key design aspects regarding the source coding part of iterative source-channel decoding (ISCD) are the index assignment and the quanizer. While the conventional index assignments are not suited for ISCD the so far presented optimized index assignments do only consider zeroth order a priori knowledge or optimize the parameter SNR only indirectly. In this paper we present a new cost function which directly optimizes the parameter SNR incorporating the first order a priori knowledge. With the same cost function we can also optimize the code book of the quantizer. Simulation results show the excellent performance the new parameter SNR optimized index assignments and quantizers exhibit.