水声通信低复杂度最大似然联合均衡译码方法

现有Turbo均衡采用迭代联合均衡译码结构,这一迭代过程在信噪比高于某一门限后才能产生正向的迭代增益,当信道码间干扰较为严重或者采用高阶调制方式时,该信噪比门限较高;同时,现有Turbo均衡的性能还受限于其交织器长度,故当码块长度较短时,其性能并不理想。针对以上问题,提出了一种基于Spinal码的联合均衡译码方法。首先,该方法采用非迭代的最大似然联合均衡译码结构,有效避免了Turbo均衡迭代结构所带来的信噪比门限高问题;其次,该方法采用信息混合能力更强的Spinal码,有助于提高码块长度受限时的误码率性能;最后,该方法利用Spinal码可以有效增加相似信息序列间汉明距离的特点,将联合均衡译码时间复杂度降低为仅随信息序列长度与信道响应长度呈线性增长,解决了本方法难以实际应用于实时水声通信的难题。在抚仙湖湖试中,同样实现速率为3.34kbps数据通信,该方法的误码率优于Turbo均衡方法两个数量级以上。时变水声信道下的仿真以及湖试结果表明,该方法在码长受限的情况下仍在复杂水声信道中表现出更好性能,其所需的最低输入信噪比显著低于现有Turbo均衡方法。

Joint equalization and decoding scheme for underwater acoustic communications using spinal code

InterSymbol Interference (ISI) cancellation over UnderWater Acoustic (UWA) channel can be performed efficiently using iterative joint equalization and decoding scheme which is also called Turbo equalization. However, the limitation of interleaver length would lead a performance loss of Turbo equalization, thus limit the performance of Turbo equalization using short block length. What's more, the iterative structure of Turbo equalization needs a high Signal-Noise-Ratio (SNR) threshold, which may also yield dramatic performance loss. We propose a joint equalization and decoding scheme to overcome these problems. Firstly, it uses a noniterative maximum likelihood joint equalization and decoding scheme to avoid the high SNR threshold. Secondly, the Spinal code with powerful ability of message-mixing is used to improve the performance under short block length. Thirdly, a linear time complexity joint equalization and decoding fast algorithm achieved. This benefit from the property of Spinal code, which increases the Hamming distance between similar source sequences. In the experiment over Fuxian Lake, the bit error rate of proposed scheme is lower than the Turbo equalization by two orders of magnitude, achieving the 3.34 kbps communication similarly. The results of time-varying UWA channel simulations and lake experiment show that, the proposed scheme outperforms Turboequalization with the same code block length. To achieve the same bit error rate,the minimum SNR needed by proposed scheme is lower than Turbo equalization.