基于时频特征的通信信号与声呐信号分类方法

Method for classifying communication signal and sonar signal based on time-frequency features

文中选取了常见的水下通信信号和三类主动声呐信号作为分类研究对象,提出了一种基于时频特征提取的水下通信信号和声呐信号的分类方法。首先对目标信号进行短时傅里叶变换,并对目标信号的时频结果进行处理得到各短时信号的瞬时频率和瞬时频率差值;然后,对目标信号的时频变化和时频差值变化曲线进行多项式拟合,提取多项式的拟合系数作为分类特征;最后,分别使用K最近邻(K-Nearest Neighbor,KNN)、支持向量机(Support Vector Machine,SVM)和反向传播(Back Propagation,BP)神经网络对四类信号进行分类识别,并对比三种方法下的分类准确度。实验结果验证了所提分类算法的有效性和可实现性。

In this paper,the common underwater communication signals and three kinds of active sonar signals are selected as the research objects,and a classification method of underwater communication signals and sonar signals based on time-frequency feature extraction is proposed.Firstly,the short-time Fourier transform of the target signal is used to get the signal time-frequency map,and the time-frequency results of the target signal are processed to get the center frequency and center frequency difference of each short-time signal.Then,the polynomial fitting of the time-frequency curve and time-frequency difference curve of the target signal is used to extract the fitting coefficient of the polynomial as the classification feature.finally,the KNN,SVM and back propagation(BP)neural network are used to identify four kinds of signals,and compares the classification accuracy of the three methods.Experimental results verify the effectiveness and feasibility of the proposed classification algorithm.