基于宽带时频图的多路摩尔斯信号自动检测

Automatic Detection of Multiple Morse Signals Based on Broadband Time-Frequency Diagram

针对现有算法没有涉及的宽带环境下多路摩尔斯信号自动检测问题,提出一种基于宽带时频图和集成学习分类器的算法.首先,通过提出一种基于宽带时频图的信号快速窄带滤波方法,实现噪声背景中各类型信号窄带时频图的快速获取;然后,为从上述窄带时频图中识别出多路摩尔斯信号,提出3个新特征与局部二值模式特征构成特征向量;最后,采用集成学习算法设计分类器实现摩尔斯信号的自动检测.与现有算法对比实验结果表明:针对多组实际数据,该算法的正确率均可达95%以上,同时误检率低于10%,具有良好的鲁棒性和应用价值.

Current algorithms don,t cover the multiple Morse signals detection in broadband environment. To solve this problem,an algorithm based on broadband time-frequency diagram and ensemble learning classifier was proposed. Firstly, a fast narrow band signal filtering method based on broadband time- frequency diagram was proposed to realize the fast filtering of narrowband time-frequency signals of various types of signals in the noise background. Then, in order to identify the Morse signals, a new feature vector was proposed which was combined with three new features and local binary pattern （ LBP）. Finally the ensemble learning algorithm was used to design the classifier to realize the automatic detection of Morse signals. The experimental results show that the correctness of the algorithm is above 9 5 % and the error rate is below 10%,which has good robustness and application value.