基于谱相关和神经网络的信号调制识别

Modulation recognition based on spectral correlation and neural network

由于通信信号的体制及调制及调制方式的复杂多样,通信信号调制类型的识别显得尤为重要和迫切。基于调制信号的谱相关特征,提取了5个特征参数,给出了各个参数随信噪比变化的曲线图。分类器采用RBF神经网络,并从提高网络识别性能出发,构建了大容量和高质量的网络训练样本,能够扩大识别范围,提高识别精度。基于谱相关特征参数和RBF神经网络结合的算法能动态识别信号的调制方式,仿真结果表明:该算法在低信噪比下能取得较高的正确识别概率。

Because of the complexity of communication signal system and modulation, the recognition of communication signal modulation type becomes particularly important. Based on the spectral correlation features of modulations, the five characteristic parameters were extracted, and the graphs of each parameter changing with signal noise ratio were presented. The RBF neural network was used as a classifier, and according to the recognition performance, the training swatch with large capacity and high quality was established so as to expand the range of recognition and improve the recognition precision. The modulation types can be identified dynamically based on spectral correlation characteristic parameters and RBF neural network. Simulation result shows that this algorithm can improve recognition probability in low SNR environment.