湖底回波的包络特征提取

Envelop Feature Extraction of Lake Bottom Echoes

湖底回波包络包含了湖底沉积物的结构和物理性质的信息,可以作为沉积物分类的特征。用传统的H ilbert变换提取宽带回波包络存在一些固有的缺点。该文中,采用线性相位的双正交小波,对湖底回波解析信号的实部和虚部分别进行离散正交小波变换,提取合适尺度上的小波系数的模值作为包络特征矢量。它可以采用M a llat快速算法,运算量少,提取的包络特征矢量维数少,能简化目标识别的算法。对实测的湖底回波数据进行特征提取和分类的仿真实验也表明,采用这种方法得到的包络特征是一种稳健、有效的特征,能获得较高的正确识别率。

Since the backscatter echoes envelop implies the structure and physical character of the lake bottom sediments, it can be taken as the feature vectors to recognize the kind of the lake bottom sediments. There are some inherence disadvantages to extract the wide - band echo envelop with the Hilbert transform. A new method is proposed that the real and imaginary parts of the analy／ic signals of the echoes are transformed with linear phase bi - orthogonal wavelet respectively, and then the modular values of the wavelet coefficients at the proper scale are taken as the envelop feature vector. It can reduce the computation load and the dimensions of the envelop feature vector with the mallat algorithm. It is also shown by the simulation experiment of the echo data from the lake trial that the envelop feature acquired with this method is a kind of robust and effective feature to recognize the kind of lake bottom sediments, and high recognition rate can be obtain.