中文摘要：

在电台识别问题的研究中,电台特征提取是识别中的重要内容,直接影响到分类器的设计和识别效果,如何在低信噪比环境下实现对信号的特征提取成为电台识别的难题。针对上述问题,利用分形维数对噪声不敏感的特性,将信号瞬时频率的分形维数作为个体特征,提出了一种新的电台识别方法。首先利用经验包络法提取信号的瞬时频率,通过盒维数和信息维数定量描述瞬时频率的复杂度,再将两种分形维数构成特征向量,最后利用最邻近分类器实现电台的分类识别。通过对4部同型号的超短波电台信号的特征提取和分类识别实验,结果表明在3dB的信噪比环境下,改进方法进行识别的正确率在96%以上,实验结果验证了所提方法的有效性。

英文摘要：

Feature extraction is one of the important research topics in individual transmitter identification, which affects directly the classifier design and the recognition performance. How to extract feature under a low signal noise ratio（SNR） is one of the hot but difficult topics within relevant studies. To solve this problem, we utilize the property that fractal dimensions are insensitive to noise to take the fractal dimensions of instantaneous frequency as the signal features and a novel approach for transmitter identification is proposed. Instantaneous frequency is extracted by the empirical envelope method and its complexity can be quantitatively described by box dimension and information di- mension which are selected as the eigenvector. Then the transmitters are identified by the nearest neighbor classifier. Experimental results of fractal feature extraction and recognition of 4 ultra-short wave transmitters show that the recognition accuracy is over 96% when SNR is 3dB, which proves that the proposed method is effective.