基于多分类器的通信信号调制方式识别

Modulation Recognition of Communication Signals with Multi-classifiers

为在恶劣接收条件下识别通信信号的调制样式 ,并实现识别范围的快速调整 ,运用决策层信息融合技术提出了一种基于多分类器的调制样式识别方法。它采用了常规特征以及三种可以快速建立的子分类器 ,其中“超盒”分类器用于界定各调制样式不同特征的取值范围 ,最近邻分类器用于分类 ,马氏距离分类器用于提高识别结果的可靠程度。该方法可以自动地在短时间内根据设定的识别范围建立对应的识别器 ,对另类调制样式或噪声有高的拒识能力。

In order to reliably recognize various modulation types of communication signals under severe conditions and quickly adjust the scope of the recognizer when necessary, a recognition scheme based on data fusion technology in decision level was proposed. The scheme employed 20 simple features and consisted of three sub-classifiers: the 'Hyper Box' classifier was designed to isolate the region occupied by each modulation type in the feature space, the nearest-neighbor classifier was as the main classifier, and the Mahalanobis classifier was used to improve the reliability of the recognition results. All the sub-classifiers could be set up instantly. Experiments with off-the-air signals show that the proposed scheme can reliably recognize concerned modulation types and meanwhile reject unwanted ones with a high probability. It can adapt to the change of recognition scope automatically and quickly.