Joint Estimation of Frame Synchronization and LDPC Encoder Using Deep Learning

Abstract

In contemporary wireless communication systems, frame synchronization and channel encoder identification are critical for reliable signal processing, particularly in scenarios with limited or no prior knowledge of the transmitted signal structure. Traditionally, these tasks are performed separately, often resulting in increased system complexity and reduced efficiency. This paper proposes a novel deep learning-based approach using capsule networks (CapsNet) to jointly estimate frame synchronization points and low-density parity-check (LDPC) encoders over Rayleigh fading channels. The model is evaluated using five different LDPC codes based on the fifth generation (5G) new radio (NR) standard and eleven distinct frame synchronization positions. The proposed model achieves over 95% classification accuracy at a 5 dB signal-To-noise ratio (SNR) under Rayleigh fading conditions for input length of 1680. Furthermore, it reaches 100% accuracy for all input lengths when the SNR exceeds 10 dB under the same channel conditions. Moreover, comparisons were made between our proposed CapsNet model and existing models from the literature, clearly demonstrating that our model significantly outperforms them in classification accuracy, especially at low SNR level. These results highlight the robustness and effectiveness of our model, demonstrating its superior performance in practical wireless communication scenarios. © 2025 IEEE.