基于瞬时特征和BP神经网络的数字调制信号自动识别及实现

Automatic Recognition and Realization of Digital Modulation Signal based on Instantaneous Features and BP Neural Network

在非协作通信环境下,针对2ASK、4ASK、2FSK、4FSK、BPSK、QPSK六种常规数字调制信号,采用了基于瞬时统计特征的信号自动分类算法对上述六种信号进行分类识别。仿真结果表明,该算法在信噪比大于15dB时,总体识别率高于90%;随着信噪比减小,信号识别率下降明显。当信噪比低于5dB时,信号识别率不足60%。为提高低信噪比下信号识别率,进一步引入BP神经网络作为分类器,采用自适应学习速率梯度下降法训练神经网络,设计了基于BP神经网络的数字调制信号自动识别方案。最后在软件无线电设备NI-USRP 2920上实现了该方案,并验证了性能。实测结果表明,六种数字调制信号正确识别率均高于91%。

In a non-cooperative communication environment,the six conventional digital modulation signals including 2ASK,4ASK,2FSK,4FSK,BPSK and QPSK are classified and identified by the automatic signal classification algorithm based on instantaneous statistical characteristics.Simulation results indicate that the overall recognition rate is higher than 90%when SNR(signal to noise ratio)is greater than 15dB.With the decrease of SNR,the recognition rate of the signal decreases obviously.When the SNR is lower than 5dB,the signal recognition rate is less than 60%.In order to improve the signal recognition rate under low SNR,the BP neural network is further used as a classifier to train the neural network through the adaptive learning rate gradient descent method to recognize six kinds of modulated signals.Finally,the performance of the algorithm is verified on the USRP platform.The measured results show that the correct recognition rate of the six digital modulation signals is higher than 90%.