基于支持向量机的分级调制识别方法

Hierarchical digital modulation recognition using support vector machines

目前大部分调制识别方法存在计算量过大和分类器训练困难等问题。针对这一现状,提出了一种基于支持向量机(SVM)的分级调制识别新方法。将接收信号的累积量和瞬时频率统计量作为分类特征参数,并利用支持向量机作为分类器对其进行分级调制分类。该方法相比其他非分级调制识别方法具有较低的计算复杂度和较快的分类器训练速度,并且对于载波频率偏移、相位抖动以及G auss噪声均具有良好的鲁棒性。计算机仿真表明,针对A SK、FSK、PSK、QAM等11种数字调制信号,当噪声采用G auss白噪声,并且信噪比≥5 dB时,正确识别率高于95%。

Most current digital modulation recognition methods are limited by the eomputational complexity and the training of the classifiers. A hierarchical method for multi class digital modulation recognition using support vector machines （SVM） was developed as a simplified method. The method uses cumulants and simultaneous frequency statistieal moments of the received signals as the features and support vector machines as the classifiers. This hierarchical method is less complex computationally and has faster classifier training speed compared with other methods. Moreover, the method is robust in the presence of carrier phase and frequency offsets with Gaussian noise. Classifieation results for 11 modulation types including ASK, FSK, PSK, and QAM obtained from computer simulations, show that the overall success rate is 〉 95% with Gaussian noise having SNR ≥ 5 dB.