Verification-Based Decoding for Rateless Codes in the Presence of Errors and Erasures

In this paper, verification-based decoding is proposed for the correction and filling-in of lost/erased packets for multicast service in data networks, which employs Rateless codes. Patterns of preferred parity-check equations are presented for the reduction of the average number of parity-check symbols required. Since the locations of unverified symbols are known, the effect of erasures and errors is the same in terms of the overhead required for successful decoding. Simulation results show that for an error-only, an erasure-only or a combination of both at 10% error/erasure probability, 78% of the messages can be recovered with a 50% overhead, whereas 99% of the messages can be recovered with a 100% overhead.