基于SVD与数学形态学分形维数谱的战场声特征提取

Battlefield Acoustic Feature Extraction Based on SVD and Math Morphology Fractal Dimensions Spectrum

分形维数作为战场声信号的特征,存在特征数量不足,反映信号非线性不充分的问题,提出了一种基于SVD与数学形态学分形维数谱(Singular Value Decomposition And Mathematical Morphological Fractal Dimensions Spectrum,SVD-MMFDS)的战场声特征提取方法。对声信号构造Hankel矩阵,再进行SVD分解,根据信号频率与奇异值的关系,重构信号分量。将这些重构信号依次线性叠加,每叠加一次信号分量就计算一次分形维数,直至完全恢复原信号;通过这种方法,构成数量多且更能反映信号非线性的分形维数谱。运用半实物仿真实验将SVD与数学形态学分形维数谱的方法,与变分模态分解(VMD)和分形维数结合的方法进行对比,该方法提取的战场声特征具有更好的区分度且特征数量更多,为利用信号非线性来识别战场声目标提供较好的选择。

When the fractal dimension is used as the features of the battlefield acoustic signal,there are not enough features and the nolinearity of the signal could not be shown sufficiently.A method to extract features based on SVD and mathematical morphology fractal dimensions spectrum(SVD-MMFDS)is proposed.Firstly,the Hankel matrix from the signal is established,and SVD decomposition is performed.Next,the signal component is reconstructed according to the relation between the signal frequency and SVD.The linear superposition of the reconstructed signal is made in turns.As a signal conponent is super imposed once,the fractal dimension is calculated until the original signal is recovered completely.By this method,the quantity of components is large and the nonlinear fractal dimension spectrum of the signals can be reflected better.Compared with SVD,VMD and mathematical morphological fractal dimensions and fractal dimension method,the simulation results showed the features extracted by this method were more with better discrimination degree.Therefore,the proposed method can be a better choice for the identification of battlefield acoustic targets with the nonlinearity of the signals.