1: \begin{abstract}

2: In this paper we investigate stopping criteria for iterative decoding from a mutual information perspective. We introduce new iteration stopping rules based on an approximation of

3: the mutual information between encoded bits and decoder

4: soft output. The first type stopping rule sets a threshold value directly on the approximated mutual information for terminating decoding. The threshold can be adjusted according to the expected bit error rate. The second one adopts a strategy similar to that of the well known cross-entropy stopping rule by applying a fixed threshold on the ratio of a simple metric obtained after each iteration over that of the first iteration. \comment{The convergence properties of

5: this mutual information are examined and }Compared with several well

6: known stopping rules, the new methods achieve higher efficiency.

7: 

8: \end{abstract}