基于VMD和改进CNN的舰船辐射噪声识别方法

Recognition method of ship radiated noise based on VMD and improved CNN

针对海上低信噪比舰船目标的识别问题,对传统卷积神经网络进行改进并与变分模态分解相结合,提出了基于VMD和改进CNN的舰船辐射噪声识别方法。应用所提方法对东海试验中12艘辐射噪声信噪比低于5 dB的舰船目标进行了识别,平均正确率为98.6%;相比于其他7种识别方法,分别提升了24.8%、17.0%、15.1%、8.0%、13.1%、16.8%、5.2%;改进卷积网络较传统卷积网络在运算量和识别速率方面有明显优势。

Here, aiming at the problem of ship target recognition with low signal-to-noise ratio(SNR) at sea, the traditional convolutional neural network(CNN) was improved and combined with the variational mode decomposition(VMD), a ship radiated noise recognition method based on VMD and improved CNN was proposed. The proposed method was applied to identify 12 ships with radiated noise SNR lower than 5 dB in the east China sea tests. The results showed that the average recognition accuracy of the proposed method is 98.6%;compared with the other 7 recognition methods, the recognition accuracy of the proposed method increases by 24.8%, 17.0%, 15.1%, 8.0%, 13.1%, 16.8% and 5.2%, respectively;compared with traditional CNN, the improved CNN has obvious advantages in computation amount and recognition rate.