基于小波包分析和深度学习的舰船辐射噪声识别

Ship Radiated Noise Recognition Based on Wavelet Packet Analysis and Deep Learning

为提高舰船辐射噪声识别的准确率,针对辐射噪声这种非平稳、复杂的信号,提出一种基于小波包分解与多特征融合的特征提取方法。同时,引入深度学习模型,将提取到的特征作为识别分类的依据,采用卷积神经网络和长短时记忆神经网络作为分类器。对单一特征的分类结果与融合的多特征分类结果进行比较,对直接提取的特征分类结果与基于小波包分解提取的特征分类结果进行比较,对卷积神经网络、长短时记忆神经网络和机器学习的识别分类结果进行比较,结果表明,采用基于小波包分解与特征融合的特征提取方法和基于深度学习的分类识别方法能显著提高舰船辐射噪声识别的准确率。

In order to improve the recognition accuracy, this paper proposes a feature extraction method based on wavelet packet decomposition and multi-feature fusion for radiated noise, which is a non-stationary and complex signal. A deep learning model is introduced to use the extracted features as the basis for recognition and classification. Convolutional neural networks and long short-term memory neural networks are used as classifier. Single-feature classification results and fused multi-feature classification results, directly extracted feature classification results and feature classification results based on wavelet packet decomposition, and convolutional neural networks recognition classification results, long and short-term memory neural networks recognition classification results and machine learning recognition classification results are compared. The results show that the feature extraction method based on wavelet packet decomposition, and the classification recognition method based on deep learning have a significant improvement of the accuracy of ship radiated noise recognition.