基于LCNN-Ⅱ模型的光通信信号调制格式识别方法

Optical communication signal modulation format identification based on the LCNN-Ⅱ model

针对光通信信号调制格式识别准确率较低的问题,提出了一种基于改进Inception结构的轻量级卷积神经网络(Lightweight Convolutional Neural Network with Improved Inception,LCNN-Ⅱ)模型的光通信信号调制格式识别方法。以不同传输条件下调制格式的星座图为基础,使用二值化、灰度增强、邻域灰度增强这3种特征图像生成算法对星座图进行处理,生成特征图像矩阵,构建调制信号的特征参数的数据集。利用改进的Inception结构,设计集成Inception模块的LCNN-Ⅱ模型,以提高模型的识别准确率。仿真结果表明,所提方法可以较好地保存光通信信息图像的特征,在传输距离为120 km的条件下,识别准确率达到了95.71%。与相关方法相比,所提方法的识别准确率较高。

For the problem of low identification accuracy when recognizing the modulation formats of optical communication signals,a modulation format recognition method for optical communication signals based on a lightweight convolutional neural network with improved Inception(LCNN-Ⅱ)model is proposed.Based on the constellation diagrams of modulation formats under different transmission conditions,constellation diagrams are processed by three feature image generation algorithms,namely binarization,grayscale enhanced,and neighborhood grayscale enhanced,to generate feature image matrices and construct the dataset of feature parameters of the modulated signals.The integrated inception module is designed to build the LCNN-Ⅱ model to improve the recognition accuracy.Simulation results show that the proposed method can effectively preserve the features of optical communication information images,and the recognition accuracy reaches 95.71% under the condition of a transmission distance of 120 km.Compared with relevant methods,the proposed method has a higher recognition accuracy.