摘要：为了提高船舶辐射噪声信号的正确识别率，本文通过小波包对船舶辐射噪声进行 分解，提取了各频段的能量谱系数作为识别特征并且结合 BP 神经网络分类器对目标 信号进行分类识别。首先介绍了小波包变换以及 BP 神经网络算法的基本原理，然后 针对 3 类不同的实测船舶辐射噪声信号进行多层小波包分解，提取分解后的各子频带 能量谱系数，经过归一化处理后，作为不同船型的识别特征，最后利用 BP 神经网络 分类器进行分类实验。

仿真结果表明，针对船舶辐射噪声信号的非线性和非平稳特征，提取的船舶辐射 信号小波包能量谱特征能反应其特性，并且 BP 神经网络分类器对 3 类船舶辐射噪声 信号具有较好的识别效果，能够很好的区分不同类型的目标信号。 

关键词：小波包分解；小波包能量谱；特征提取；BP 神经网络分类器；目标识别

Abstract：In order to improve the correct recognition rate of the ship-radiated noise signal, this thesis extracts the energy spectrum coefficient of each frequency as the recognition feature incorporated with BP neural network classifier to classify the target signal through decomposition from wavelet packet of the radiated noise of ship. Firstly, the basic principle of wavelet packet transform and BP neural network algorithm is introduced. Secondly, through the multi-layer wavelet packet decomposition of three different types of ship-radiated noise signals, the energy spectrum coefficients of each frequency  are extracted as the identification characteristics of different ship types after normalization. At last, classification experiments are performed utilizing BP neural network classifier.

The simulation results show preferable performance in the wavelet packet energy spectrum of the ship-radiated noise signal, extracted from the characteristics of nonlinear and nonstationary and the recognition results, which are able to distinguish different types of target signals, of BP neural network classifier on three types of ship-radiated noise signals.

Keywords：wavelet packet decomposition; wavelet packet energy spectrum;   feature extraction; BP neural network classifier; target recognition

目 录

第一章 绪论 1

1.1 研究背景与意义 1

1.2 基于船舶辐射噪声小波包能量谱特征的船型识别研究现状与发展 1

1.2.1 小波包能量谱特征的研究现状和发展 1

1.2.2 BP 神经网络分类器的研究现状和发展 2

1.3 论文的主要研究内容和内容安排 3

第二章 基于船舶辐射噪声小波包能量谱的特征提取 4

2.1 概述 4

2.2 船舶辐射噪声产生机理 4

2.3  小波包分析理论 5

2.3.1 小波分析 5

2.3.2 最佳小波基函数选取 5

2.3.3 多分辨分析和 Mallat 算法 6

2.3.4 小波包分析 7

2.4  船舶辐射噪声小波包能量谱特征提取 9

2.5  本章小结 10

第三章 BP 神经网络的原理及方法