Design of Quantized Decoders for Polar Codes using the Information Bottleneck Method

Abstract:
Polar codes are a relatively new introduction to the family of forward error correcting codes. This work deals with reducing the implementation complexity of polar decoders. To that end, a discrete successive cancellation list decoder is developed which mostly processes only unsigned integers instead of LLR values. Further, all operations inside the discrete decoder are simple table lookups. These two modifications lead to reduction in the complexity of the decoder. We show that the block error rate performance of the proposed discrete decoder is very close to that of the conventional successive cancellation list decoder.