On Optimized Uniform Quantization for SC Decoder of Polar Codes

Polar codes have a regular recursive structure which is feasible for hardware implementation in practice. Quantization is a critical issue in hardware implementation. In this paper, a quantization scheme for successive-cancellation (SC) decoding of polar codes is proposed based on the equivalent channel cutoff rate optimization. Moreover, a modified Gaussian approximation (GA) method exploiting the cutoff rate maximizing criterion is brought forward to construct polar codes and estimate frame error rate (FER) performance under the quantized decoding algorithms. Utilizing the modified GA, the optimal quantization bits can be determined. Simulation results demonstrate that a 6-bit uniform quantized SC decoder could achieve a near floating point performance. Compared with the existing results in the published literature, the proposed method consumes no more than 4.5 quantization bits and performs better.