Space-Efficient Quantized Polar Decoders Designed using the Information Bottleneck Method

Inhalt:
Quantized successive cancellation list decoders can be designed by incorporating the information bottleneck method into the density evolution of polar codes. All operations, except the path metric update, in the information bottleneck successive cancellation list decoders are simple table lookups of unsigned integers. The number of lookup tables required for decoding depends on the codeword length such that a larger number of tables are used in decoding longer codewords. These lookup tables add to the space complexity of the quantized decoder and it is, therefore, desirable to reduce the number of lookup tables required for decoding. In this work, an information bottleneck based polar decoder design is proposed that uses the minsum rule to design the lookup tables beside the information bottleneck method. In the proposed decoder, the number of required decoding tables is halved since the min-sum lookup tables need not to be stored due the availability of their hardware-efficient implementation. Additionally, the design SNR of the information bottleneck decoders are optimized to improve their error correction performance.