利用遗传BP神经网络的调制识别新算法

Novel Modulation Recognition Algorithm based on Genetic BP Neutral Network

在非合作通信领域中通信信号调制方式的自动识别具有重要作用,如何在低信噪比下准确识别接收到的信号是这一领域研究的重点。针对这种情况,利用信号高阶累积量、瞬时幅度谱以及信号N次方非线性变换后的特征提取三个新的特征参数,并采用遗传算法优化的BP神经网络作为分类器,提出一种利用遗传BP神经网络的信号调制识别算法。仿真识别2FSK、BPSK、QPSK、UQPSK、8PSK五种通信中常用的调制信号。BPSK、QPSK信号在0dB时识别率可达到96%以上,其余信号在信噪比大于0dB时识别率均能达到85%以上。实验表明该算法在低信噪比下对上述信号具有良好的识别效果。

In the field of non-cooperative communication,the automatic identification of communication signal modulation methods plays an important role.How to accurately identify the received signal at low SNR is the focus of research in this field.Aiming at this situation,a new signal modulation recognition algorithm is proposed by using the classifier of BP neural network optimized by genetic algorithm and the three new characteristic parameters extracted by using the high-order cumulants of signals,the instantaneous amplitude spectrum and the characteristics of the N-th power nonlinear transformation of the signals,which is used to identify 2FSK,BPSK,QPSK,UQPSK and 8PSK signals.The recognition rate of BPSK and QPSK signals could reach 96% at 0dB,and the recognition rate of the other signals could reach 85% when the SNR is not lower than 0dB.Experiments indicate that the algorithm has a good performance in recognizing the above signals at low SNR.