基于Wavelet-CNN的电磁炮过靶信号识别方法

电磁炮测试中,炮口产生强烈的火光信号以及振动等噪声,会严重干扰电枢特征信号的识别处理。为了提升对电枢信号的自动识别率,提出了一种基于小波变换和卷积神经网络(Convolutional Neural Network,CNN)相结合的电枢信号识别方法。首先,利用小波变换对过靶信号进行小波阈值去噪,进而重构信号。其次,利用CNN提取信号的深层次特征,通过CNN的全连接层输出信号的分类结果。最后,当输入信号为电枢信号时,对其作最大值检测获取电枢信号的特征点。实验结果表明,本文所提方法对比传统小波阈值滤波法在特征点自动拾取准确率上提升了5.88%。该算法对电磁炮电枢过靶信号的滤波、识别具有一定的参考意义。

Wavelet-CNN-based Identification Method For Electromagnetic Gun Over-target Signals

In electromagnetic gun testing, the muzzle generates strong fire signals and noise such as vibration, which can seriously interfere with the recognition processing of armature feature signals. In order to improve the automatic recognition rate of armature signals, an armature signal recognition method based on the combination of wavelet transform and Convolutional Neural Network (CNN) is proposed. Firstly, wavelet transform is used to denoise the over-target signal with wavelet threshold, and then reconstruct the signal. Secondly, the deep features of the signal are extracted using CNN, and the classification results of the signal are output through the fully connected layer of CNN. Finally, when the input signal is an armature signal, the maximum detection is performed to obtain the feature points of the armature signal. The experimental results show that the proposed method improves 5.88% in feature point automatic picking accuracy compared with the traditional wavelet threshold filtering method. The algorithm has certain reference significance for the filtering and identification of electromagnetic gun armature over-target signals.