基于YOLO-OpenMax的水声通信信号开集识别方法

Open-Set Recognition of Underwater Acoustic Communication Signal Based on YOLO-OpenMax

水声通信信号调制方式随着技术进步从典型的传统调制方式发展为水声通信机厂商私有开发的新型调制方式,这使得现有的水声通信信号闭集调制识别技术实用性受到制约,但现阶段鲜有研究基于深度学习的水声通信信号调制方式的开集识别方法。针对此现状,提出一种基于YOLO图像检测识别网络和OpenMax模型的开集识别方法。利用YOLOv5网络的输出特点,将传统OpenMax模型的处理方法进行改进,提出一种两步测试方法,实现了在多径信道下的水声通信信号的开集识别。仿真实验表明:在归一化系数为0.5,信噪比等于10 dB条件下的归一化开集识别准确率达到90%以上,实测数据同样验证了本文方法的有效性。

As technology develops,the modulation type of underwater acoustic communication signals has gradually evolves from typical traditional modulation methods,to many new modulation methods developed by underwater acoustic communication machine manufacturers privately,which restricts the practicality of the existing closed-set modulation and identification technology of underwater acoustic communication signals.To address the problem that there are few open-set recognition methods based on deep learning for underwater acoustic communication signals,an open-set recognition method based on the YOLO network and OpenMax model was proposed.Using the output characteristics of the YOLOv5 network,the traditional process of the OpenMax model was improved,and a two-step testing process was proposed to realize the open-set recognition of underwater acoustic communication signals under multipath channels.Simulation experiments show that when the normalization coefficient is 0.5 and the signal-to-noise ratio is 10 dB,the normalized accuracy reaches more than 90%.The measured data also verify the effectiveness of the proposed method.