莱斯衰落信道下 MQAM 信号调制方式自动识别方法

针对受莱斯衰落影响的4QAM、16QAM、32QAM、64QAM、128QAM、256QAM六类信号，分别研究了卷积神经网络（CNN）模型以及特征参数结合深度神经网络（DNN）分类器模型的调制方式识别性能。CNN模型需要大量带标签的数据集以及很长的训练时间才能获得较好的识别性能，而特征参数结合深度神经网络分类器模型所需训练时间较短，但其分类性能受限于特征参数的设计。针对以上问题，研究了混合高阶矩作为特征参数集，再将DNN作为分类器对多进制正交幅度调制（MQAM）信号进行识别的方法。仿真结果表明，该方法在低信噪比情况下对受莱斯衰落影响的MQAM信号识别准确率高于CNN模型，且分类准确率上限明显高于采用高阶累积量作为特征参数的方法。

Automatic Recognition of MQAM Signals Modulation over Rician Fading Channel

The modulation recognition performance of different models,convolutional neural network (CNN) and designed feature parameters combined with deep neural network (DNN) classifier were studied respectively for identifying six types of signals affected by Rician fading which are 4QAM, 16QAM, 32QAM, 64QAM, 128QAM and 256QAM. A large number of labeled datasets and long-time training are needed for CNN model to obtain satisfying recognition performance. The model which combines designed feature parameters with DNN classifier requires shorter training time, but the classification accuracy of the model is also limited by the designed parameters. In view of the above situation, the mixed high order moments of MQAM signals were studied to construct characteristic parameters, and then DNN was used as a classifier to identify various multiple quadrature amplitude modulation (MQAM) signals. Simulation results show that the recognition accuracy of MQAM signals affected by Rician fading is higher than that of CNN model under the condition of low signal to noise ratios (SNRs), and the upper limit of classification accuracy is apparently higher than that of using high order cumulants as characteristic parameters.