Decoding for Punctured Convolutional and Turbo Codes: A Deep Learning Solution for Protocols Compliance

TL;DR

The paper tackles the challenge of neural decoders for punctured convolutional and Turbo codes, which must adapt to multiple code rates while remaining protocol-compatible. It introduces a unified LSTM-based Convolutional Neural Engine (CNE) that incorporates a puncturing-aware embedding, enabling seamless generalization across protocol-driven rate patterns, and a balanced BER training scheme to stabilize learning across rates. Through extensive AWGN and Rayleigh-channel experiments, the CNE demonstrates state-of-the-art decoding performance, including a $0.2$ dB gain at BER $10^{-4}$ under matched conditions and substantial gains in Rayleigh scenarios, often surpassing traditional decoders even with imperfect CSI. The work further discusses complexity and latency considerations, offering practical strategies for deployment on hardware accelerators and outlining avenues for future extension to broader coding schemes while preserving protocol compatibility.

Abstract

Neural network-based decoding methods show promise in enhancing error correction performance but face challenges with punctured codes. In particular, existing methods struggle to adapt to variable code rates or meet protocol compatibility requirements. This paper proposes a unified long short-term memory (LSTM)-based neural decoder for punctured convolutional and Turbo codes to address these challenges. The key component of the proposed LSTM-based neural decoder is puncturing-aware embedding, which integrates puncturing patterns directly into the neural network to enable seamless adaptation to different code rates. Moreover, a balanced bit error rate training strategy is designed to ensure the decoder's robustness across various code lengths, rates, and channels. In this way, the protocol compatibility requirement can be realized. Extensive simulations in both additive white Gaussian noise (AWGN) and Rayleigh fading channels demonstrate that the proposed neural decoder outperforms conventional decoding techniques, offering significant improvements in decoding accuracy and robustness.

Figures (16)

• Figure 1: Encoding structure of the convolutional code in IEEE 802.11.
• Figure 2: Example of the bit-stealing procedure in IEEE 802.11 (R = 2/3, 3/4, 5/6).
• Figure 3: Encoding structure of the Turbo codes in 3GPP TS 36.212.
• Figure 4: Example of the rate matching procedure in 3GPP TS 36.212.
• Figure 5: The unrolled recurrent neural network.