舰船辐射噪声的混沌特征提取方法研究

【作者】 陈静；

【导师】 李亚安；

【作者基本信息】 西北工业大学 ， 水声工程， 2006， 硕士

【摘要】 水下目标特征提取技术在现代水声信号处理领域具有十分重要的意义。将混沌理论应用于水下信号处理，通过相空间重构、混沌特征参数提取等非线性时间分析方法可以达到对舰船信号识别的目的。本文以基于混沌理论的非线性时间序列分析为基础，围绕舰船辐射噪声的特征提取这一主题，进行了如下工作： 1．在相空间重构的基础上，对Takens嵌入定理进行了研究，并对重构的两个参数—延迟时间和最小嵌入维数进行了估计。利用平均互信息法确定相空间重构的时间延迟。利用伪最邻近点法确定相空间重构的最小嵌入维数。介绍了非线性动力系统状态变量之间的信息流概念，有助于进一步了解非线性的本质。 2．研究了基于相空间重构的非线性局部投影滤波方法，该方法的原理是将低维的时间序列拓展到高维的相空间，将高维相空间分解为两个正交的子空间，将具有较小特征值的特征矢量所在的流形向较大的特征值所在的特征矢量投影。利用局部投影滤波方法对舰船辐射噪声进行降噪，可以较好地恢复波形和相空间轨迹。 3．研究了舰船辐射噪声的特征提取方法。通过对舰船辐射噪声进行关联维数、最大Lyapunov指数、h₂熵等混沌特征参数的提取，结果表明：利用关联维数的确可以对不同的舰船目标进行分类；由于舰船辐射噪声具有正的最大Lyapunov指数及正的h₂熵，说明舰船辐射噪声中确实有混沌成分存在。 4．研究了基于相空间重构的递归图及定量递归分析。通过对舰船辐射噪声递归图的分析及对定量递归分析特征量的计算，得出：舰船辐射噪声的递归图与其混沌特征量如关联维数之间存在着某种关系；定量递归分析的特征量，如确定率、递归率和熵等，可以作为混沌特征参数对实际舰船信号进行识别、分类：定量递归分析的特征量与混沌特征参数之间具有某种对应关系；提出：递归分析是一种新的非线性方法分析的有效工具。更多还原

【Abstract】 Underwater targets feature extraction has a very important effect on underwater acoustics. The ship targets recognition can be reached up by the nonlinear time series analysis method of phase reconstruction and chaos feature extraction. In this paper, some work has been done to extract the ship radiated noise signal on the basis of nonlinear methods and chaos theory, the main contribution of the dissertation is as follows:1. The Takens embedding theory is explored and two parameters of reconstruction including the delay time and the minimum embedding dimension are estimated. The delay time of reconstruction is confirmed by the average mutual information method. The minimum embedding dimension of reconstruction is confirmed by the false nearest neighbours method. The concept of information flow among the nonlinear state variable can help understanding the nonlinear essence.2. Nonlinear local project noise reduction on the basis of phase space reconstruction is studied. The principle of this method is to develop the lower time series into the higher phase space and decompose the higher into two orthogonal sub-spaces, then project the sub-space which has little eigenvalues to the eigenvectors which has larger eigenvalues. The nonlinear local project noise reduction method has been improved greatly to recover the origin signal and trajectories of phase space.3. Feature extraction of ship radiated noise signal is explored. It can be learned that correlation dimension can be used to classify the different types of underwater objects; there are positive maximum Lyapunov exponents and h₂ entropy, this indicated that ship radiated noise signals are chaotic.4. The recurrence plots and quantification recurrence analysis on the basis of state space reconstruction are explored. By analyzing the recurrence plots of ship radiated noise and calculating the parameters of quantification recurrence analysis, it can be learned that there are some connections between the recurrence plots of ship radiated noise and the chaos parameter such as correlation dimension; some parameters of quantification recurrence analysis can identify and classify the actual underwater object; recurrence analysis is a new valid nonlinear analyzing tool.更多还原