多元统计分析在舰船辐射噪声分类识别中的应用
文献类型：学位论文
| 作者 | 张岩 |
| 学位类别 | 博士 |
| 答辩日期 | 2007-07-05 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 声学研究所 |
| 关键词 | 舰船辐射噪声 多元统计分析 功率谱 Lofar谱图 主成分分析 PCA 独立成分分析 ICA 分类识别 |
| 其他题名 | The Application of The Multi-parameter Statistics Analysis on The Feature Extraction of The Ship-radiated Noises |
| 学位专业 | 信号与信息处理 |
| 中文摘要 | 有效的特征提取是目标分类识别的基础。在噪声信号特征分析中，比较传统的方法有傅立叶变换、基于谱估计的分析、时频分析等，但通常存在原始特征数据量大、维数高、计算成本高的问题。多元统计分析是近年来信号处理领域发展很快的一个分支，在处理多变量综合问题和信号混叠数据方面有比较突出的优势。如何在传统方法的基础上，将多元统计分析应用于舰船辐射噪声的信号特征提取和分类识别，是本文将要研究的问题。 本文主要探讨了两类多元统计分析方法：主成分分析（PCA）和独立成分分析（ICA）。主成分分析是多元统计分析中比较经典的方法。在形成主分量的过程中，通过舍弃小方差项、保留大方差项，减少了所需特征参数数量，为数据降维提供了有效方法。同时，基于PCA的数据重构则为信号识别提供了新的方向。独立成分分析是现代信号处理的常用方法，是解决盲源信号分离问题最有效的工具之一。该方法的一个重要前提是源信号的统计独立性，相互独立在统计学中是个很强的条件，但在现实生活中未知源信号相互独立则是可能性很大的。 本文首先通过对传统舰船辐射噪声信号特征分析方法的探讨，比较了各种功率谱估计方法，实现了线谱提取，并通过Lofar谱图探讨了舰船辐射噪声信号的时频特性；之后对PCA和ICA方法的原理和应用进行了详细研究；最后在经典功率谱和Lofar谱图的基础上尝试运用PCA和ICA的不同算法分别对两种舰船辐射噪声信号进行了特征提取和分类识别，并对结果进行了分析比较。 |
| 英文摘要 | Efficient feature extraction is the basis of the target classification and recognition. Traditional methods of noise feature analysis based on Fourier Transform, spectrum estimation, time-frequency analysis, etc. usually have problems with the highly-cost calculation due to the large amount and high dimension of the original features. The multivariate statistical analysis is a quickly developing method in the field of signal processing these years, which has obvious advantage in the integrated multi-variable problem and the mixed signal processing. This paper will study on the topic of the application of the multi-parameter statistics analysis on the feature extraction of the ship-radiated noise based on the traditional methods. This paper puts emphasis on 2 kinds of multivariate statistical analysis including “Principal Component Analysis (PCA)” and “Independent Component Analysis (ICA)”. Principal Component Analysis is a kind of classic method of multivariate statistical analysis, which reduces the necessary number of the feature parameters and provides effective method of lowering data’s dimension by means of abandoning small-variance-terms and keeping big-variance-terms. Independent Component Analysis is one of the most effective tools to solve the blind source separation problem and has been commonly used in modern signal processing. There is a primary precondition in this method that the sources must be independent one another, which is although a strong condition in statistics, yet highly possible in real world. In this paper, firstly, the traditional method of the ship-radiated noise’s feature analysis is discussed, based on which the compare of several kinds of power spectrum estimate is accomplished with the extracted line-spectrum; besides, the time-frequency feature of the ship-radiated noise is also discussed via Lofar spectrum. Then the paper traverses the theories and applications of PCA and ICA. Lastly, on the base of the classic power spectrum and Lofar spectrum, features of the noise radiated from 2 different types of ships are extracted using different PCA and ICA algorithms; the results are also compared and analyzed. |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 85 |
| 源URL | [http://159.226.59.140/handle/311008/264]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 张岩. 多元统计分析在舰船辐射噪声分类识别中的应用[D]. 声学研究所. 中国科学院声学研究所. 2007. |
入库方式： OAI收割
来源：声学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。