Channels with insertions and/or deletions appear often in practical communication systems (e.g. recording channels); however, analysis and signaling over such channels seem to be challenging. For instance, even for the simplest models of insertions/deletions, the channel capacity is unknown, and only upper and lower bounds on it are available. In terms of specific channel codes over insertion/deletion channels, the use of powerful outer codes such as low density parity check (LDPC) codes with inner synchronization codes such as watermark or marker codes is a promising approach for near capacity transmission. 

In this presention, we consider serially-concatenated coding schemes over channels impaired by insertion, deletion and substitution errors. We focus on the interleaved concatenation of an outer channel code and an inner marker code. To limit the decoding latency, marker code-based synchronization is performed only once per received packet, i.e., iterations with the outer decoder are not allowed. We first evaluate, through mutual information calculations, the information rate of this concatenated scheme when standard bit-level synchronization is performed as proposed in the earlier literature. We then introduce a novel symbol-level synchronization algorithm that works on groups of consecutive bits, and show that it improves the achievable rate. This algorithm is combined with appropriate message passing decoders to exploit the correlations in the likelihoods of the consecutive bits. In addition to the achievable rate analyses, which allow us to optimize the marker code, we also report error-rate simulation results that confirm the superiority of symbol-level synchronization.