In this paper, we consider a binary erasure channel (BEC) with an unknown erasure probability of δ at a transmitter. In addition, we consider that δ has a constant value in each transmission block and the transmitter knows the probability distribution of δ. For this problem with the infinite block length, based on the distribution of the random variable δ, a multi-layer encoding strategy at the transmitter and a successive decoding strategy at the receiver are proposed. In our proposed scheme, based on the value of δ, the receiver can decode a part of transmitted data from different transmission layers. In order to have a comparison with other erasure correction codes, we generalize the result of the infinite block length to the finite block-length transmission. We show that considering the average transmission rate, our scheme has a better performance than the other erasure correction codes such as the Raptor and the MDS codes. We also analyze both the decoding complexity and bit error rate versus the erasure probability of the channel. Besides, by using a numerical comparison, we try to show that our proposed strategy outperforms other transmission strategies.
