基于粒子滤波和希尔伯特-黄变换理论的水声信号处理研究

Based on particle filter theory and the Hilbert Huang transform theory, some research will be performed. Those include noise reduction, feature extraction and classification of underwater acoustic signal. (1) Typical characteristics of underwater acoustic signal are non-linear and non-stationary. Underwater acoustic signal is produce with the complex formation mechanism, so it is very difficult to establish an accurate mathematics model. In order to solve the problem, a modeling method of underwater acoustic signal based on genetic algorithm is proposed. By the proposed method, explicit mathematical expressions of underwater acoustic signal can be given. However, existing modeling methods do not have this ability. (2) Based on the algorithm principle about particle filter and the mathematical model of real underwater acoustic signal, the noise reduction method of underwater acoustic signal is proposed. The feasibility and validity of the proposed method are proved by the experiment of measured underwater target signals. The results will provide particle filter algorithm with their first taste of underwater acoustic field, also provide a new method for noise reduction of underwater acoustic signal. (3) Applying the improved Hilbert-Huang transform, feature extraction and classification for underwater target signals is discussed. Different types of measured underwater target signals are chosen as sample data. Some feature parameters which can reflect the essential characteristics of the signal are extracted and applied to the classification of underwater target signals. These features include (i) the center frequency of the strongest intrinsic mode function, (ii) the energy difference between the high and low frequency, (iii) the instantaneous energy variation range, (iv) time-frequency entropy, (v) the higher-order cumulant, and so on. Those may offer a good solution for automatic recognition of underwater targets.

以粒子滤波和希尔伯特-黄变换理论为基础，研究水声信号的降噪、特征提取以及利用特征参数进行水下目标的分类。1) 针对水声信号形成机理复杂难以建立准确的数学模型问题，研究基于遗传算法的水下目标信号的非线性建模方法，这将解决以往的水声信号建模方法中都未给出明确的数学表达式这一问题。2) 依据遗传算法建立的水下目标信号的非线性模型，提出基于粒子滤波的水声信号降噪方法，通过实测水声信号验证该方法的有效性，这将把粒子滤波的应用拓展到水声领域，为水声信号的降噪提供一种新思路。3) 利用改进的希尔伯特-黄变换理论对降噪后的实际水声信号进行特征提取和分类研究。以不同类别、一定数量的实测水下目标信号为样本，讨论他们的最强固有模态函数中心频率、高低频能量差、瞬时能量变换范围等特征参数的计算方法，提取能反映出水下目标信号本质特征的参数，并据此开展水下目标的分类研究，这将为水下目标的自动识别提供重要的参考价值。

水声信号处理是海洋领域乃至信息领域最为活跃的学科之一，也是未来反潜战和水声对抗装备发展中的关键技术。本课题利用粒子滤波、经验模态分解、混沌理论和混沌振子对水声信号处理进行了深入系统的研究。具体研究成果包括：（1）提出了水声信号建模方法，解决以往的水声信号建模方法中未给出明确的数学表达式这一问题。（2）提出了6种水声信号降噪方法，能够进一步消除噪声，混沌吸引子更加平滑清晰。（3）提出了3种水声信号特征提取方法，提高了舰船辐射噪声的识别能力。（4）提出了5种水声信号预测方法，取得了较高的预测精度。（5）提出了2种水声信号检测方法，提高了检测效果，降低了计算复杂度。利用现有海试数据验证了方法的有效性。该项目的研究在解决水下目标信号的降噪、特征提取、预测和检测方面取得一定进展，将提高水下典型目标（潜艇、舰艇等）的探测能力。