Abstract In this paper, we demonstrate a novel dual-branch multi-layer perceptron (DBMLP) based post-equalizer(PE). Since the two branches of the DBMLP can compensate linear and nonlinear distortion in the received signal, respectively, the complexity of the neural network is reduced and the performance of the post-equalizer is improved. Compared with multi-layer perceptron (MLP) based PE the space complexity of DBMLP based PE reduces 59.5%, and the bit error rate (BER) at 3.1 Gbps reduces 45.5%. The reduction of spatial complexity helps the application of post-equalizer based on neural network algorithms in practical communication systems. Meanwhile, increased BER performance helps underwater visible light communication (UVLC) systems reach higher rates. We achieved 3.2Gbps in the 64 quadrature amplitude modulation (64QAM)-carrierless amplitude and phased modulation (CAP) modulated UVLC system using DBMLP based PE. To the best of our knowledge, this is the highest record of data rate in the field of one single chip blue LED based UVLC system.

中文翻译：

---

基于后均衡器的双分支多层感知器3.2Gbps水下可见光通信系统

摘要 在本文中，我们展示了一种基于后均衡器（PE）的新型双分支多层感知器（DBMLP）。由于DBMLP的两个分支可以分别补偿接收信号中的线性和非线性失真，降低了神经网络的复杂度，提高了后均衡器的性能。与基于多层感知器 (MLP) 的 PE 相比，基于 DBMLP 的 PE 的空间复杂度降低了 59.5%，3.1 Gbps 的误码率 (BER) 降低了 45.5%。空间复杂度的降低有助于基于神经网络算法的后均衡器在实际通信系统中的应用。同时，提高 BER 性能有助于水下可见光通信 (UVLC) 系统达到更高的速率。我们达到了 3。使用基于 DBMLP 的 PE 的 64 位正交幅度调制 (64QAM)-无载波幅度和相位调制 (CAP) 调制 UVLC 系统中的 2Gbps。据我们所知，这是基于单芯片蓝光 LED 的 UVLC 系统领域数据速率的最高记录。