水下蓝绿光GRU接收机

GRU Receiver for Underwater Communication with Blue-green Laser Light

水下光信道严重的随机非线性时变特性导致建立准确信道模型的困难,门控循环单元(Gated Recurrent Unit,GRU)是一种优化的深度循环神经网络(DRNN),具有对时间序列信号之间关系的准确的建模能力。把GRU应用到水下光接收机中,通过训练数据优化GRU网络结构,使GRU接收机具备检测接收码元序列与对应发送码元序列对应关系的功能。同传统接收机比较,这种方法明显改善了水下蓝绿光接收机的性能。

The serious random nonlinear time-varying characteristics of underwater optical channel leads to the difficulty of establishing accurate channel model.Gated recurrent unit(GRU)is an optimized Deep Recurrent Neural Network(DRNN),which has the ability to accurately model the relationship between time series signals.The GRU is applied to the underwater optical receiver,and the GRU network structure is optimized through the training data,so that the GRU receiver has the function of detecting the corresponding relationship between the received symbol sequence and the corresponding transmitted symbol sequence.Experiments show that this method significantly improves the performance of underwater blue-green light receiver.