An Improved LT Encoding Algorithm Based on Reselecting Check Nodes

LT codes can be used for adaptive data transmission without feedback, but there exists a high error floor for LT codes in additive white Gaussian noise (AWGN) channels. Therefore, an improved encoding scheme is designed to eliminate the low-degree information nodes which caused the error floor. Firstly, the conventional scheme is used to provide an initial information degree distribution. Then, the threshold is introduced, and the information nodes are classified according to their degrees. The information nodes whose degree is less than the threshold will reconnect check nodes. Finally, all the information nodes will obtain a reliable degree not less than the threshold. Simulation results show that the proposed scheme increases the decoding success probabilities, and thereby the bit error rate (BER) performance is significantly improved. For example, the error floor can be reduced by nearly 4 orders of magnitude when the code length is 256.