空间耦合低密度奇偶校验码的深度迭代译码算法设计

Deep Iterative Decoding Algorithm for Spatially Coupled Low Density Parity Check Codes

空间耦合低密度奇偶校验(spatially coupled low density parity check,SC-LDPC)码在次最优迭代译码算法下能够达到最大后验概率(maximum a posterior,MAP)译码性能,但其优异的性能需要在码长很长迭代次数很多时才能实现。当采用传统迭代译码算法时,实现的复杂度将以指数增加,无法应用。为有效降低译码复杂度,滑窗译码算法被应用于空间耦合LDPC码的译码,但由于引入窗口截断,会造成译码性能的损失。针对上述问题,结合深度学习技术提出了一种空间耦合LDPC码的深度迭代译码算法。通过在消息传递过程中引入权重系数并采用深度神经网络对其进行训练获取权重系数,以此优化消息的可靠性度量值,从而加快译码收敛速度,提升译码性能。仿真结果表明:当传输在加性高斯白噪声信道时,所提的深度迭代译码算法在相同迭代次数下的译码性能均优于传统迭代译码算法和滑窗译码算法。

Spatially coupled low density parity check(SC-LDPC)codes have been proven to achieve the maximum a posterior(MAP)decoding performance under the iterative decoding algorithm.However,the excellent performances are realized with large codelength and more iterations.Using the traditional iterative decoding algorithm,the complexity is increased exponentially,which makes it impractical.To reduce the complexity,sliding window decoding algorithm was applied to decode the SC-LDPC codes,but it led to the performance loss due to the window truncation.To solve this problem,a deep iterative decoding algorithm for SC-LDPC codes was proposed.During the message passing,weight coefficient was introduced and then be trained using deep neural network.By this way,the reliability values were optimized,which can accelerate the decoding convergence and improve the decoding performance.Simulation results show that the proposed deep iterative decoding algorithm has better decoding performances than the traditional iterative decoding algorithm and sliding window decoding algorithm with the same iterations over the additive white Gaussian noise channels.