基于半监督聚类理论的MQAM信号的盲识别

Blind Recognition of MQAM Signals Based on Semi-supervised Clustering Theory

在MQAM信号的调制识别中,传统聚类算法聚类效果差,误差平方和函数出现起伏且收敛慢。对此问题,提出由标记的样本点来指导隶属度及聚类中心的更新的半监督聚类理论重构MQAM信号星座图的方法。通过分析星座图,提出了基于星座图圆半径的识别方法,完成了对不同阶数MQAM信号调制方式的识别。仿真结果表明该方法提高了聚类准确度,误差平方和函数曲线平滑,且MQAM信号的识别率在90%以上。

In the modulation classification of MQAM signals, traditional clustering algo- rithm results is poor, squared error function appears downs and slow convergence. To solve this problem, We proposed a method by the labeled samples to guide the membership degree and the clustering center update for reconstruction MQAM signal constella- tion based on semi supervised clustering theory. Through analysis the constellation, recognition method is proposed based on constellation radius, completed the recognition of the different orders of MQAM signal modulation. Simulation results show that this method improves the accuracy of clustering; the squared error function curve is smooth, and the recognition rate MQAM signals above 90%.