基于径向基神经网络的水声微弱信号检测  

Detection of Underwater Weak Acoustic Signal Based on Radial Basis Function Neural Network

作　　者：孟宪超[1] 刘纯武[1] 刘晔[1] 

机构地区：[1]国防科学技术大学机电工程与自动化学院,湖南长沙410073

出　　处：《鱼雷技术》2011年第5期349-354,共6页Torpedo Technology

摘　　要：为了克服传统随机信号处理手段检测水声微弱信号的缺陷,提高水声微弱信号检测效率,详细分析了海洋混响噪声的混沌特性,将混沌理论中相空间重构理论与径向基神经网络相结合建立了预测模型,从模型预测值与实际观测值的绝对误差中检测出微弱谐波信号频谱值,通过解算获取回波信号包含目标信息,从而提出了一种海洋混响混沌序列预测方法。相比传统随机信号处理手段检测水声微弱信号的方法,本文建立的预测模型克服了将背景噪声简化作白噪声的低信噪比检测缺陷,检测结果准确,检测效率高,适用于水声微弱信号检测。To enhance the detection efficiency of underwater weak acoustic signal,we analyze the chaotic characteristic of ocean reverberation,establish a prediction model by combining the theory of phase space reconstruction with the radial basis function neural network（RBF NN）,and detect the frequency spectrum of the weak harmonic signal from the absolute error of the predicted value and the observed value.From the spectrum we calculate the target information,thus present a chaotic serial prediction method of ocean reverberation.Compared with the traditional random signal processing method,our prediction model overcomes the detection shortcoming of detection with low signal to noise ratio（SNR） due to simplifying background noise to white noise,and can obtain detection result accurately and efficiently.The proposed model can be suitable for weak signal detection in underwater acoustic.

关 键 词：信号检测 混响 混沌 相空间重构 径向基神经网络 

分 类 号：TJ630[兵器科学与技术—武器系统与运用工程] TN911.7[电子电信—通信与信息系统]