基于LSTM的ZigBee网络中通信信号识别算法

Communication Signal Recognition Algorithm in ZigBee Network Based on Long and Short-Term Memory

针对ZigBee网络与WiFi网络共享频段,易受到WiFi信号干扰的问题,提出了一种基于长短期记忆神经网络的ZigBee网络中通信信号识别算法,对ZigBee网络中的ZigBee信号与WiFi信号进行识别。通过Matlab中的Simulink仿真搭建了ZigBee信号、WiFi信号发生器模型,来获取同相和正交(IQ)数据,在同一信道下进行干扰,并在分别设置-50dB-0dB不同的信噪比进行干扰,以此构建ZigBee、WiFi信号数据集,对数据采用不同数据维度,并对数据进行长度的截取处理;随后利用文中构建的长短期记忆神经网络(LSTM)识别ZigBee网络中通信信号,并与卷积神经网络(CNN)对比。实验结果表明:在低信噪比环境下,长短期记忆神经网络(LSTM)相比卷积神经网络(CNN)获得更高的识别准确率,LSTM网络的信号识别准确率可以达到90.5%,且LSTM模型始终能够获得较高的准确率和较为快速的收敛速度。

In order to solve the interference problem caused by WiFi in Zigbee network,a communication signal recognition algorithm based on Long and Short-Term Memory is proposed to identify ZigBee signal and WiFi signal in ZigBee network.By using Simulink simulation in Matlab,ZigBee signal and WiFi signal generator models were built to obtain in-phase and orthogonal(IQ)data,which were interfered in the same channel.Different signal-to-noise ratios of-50dB-OdB were set for interference,in order to construct ZigBee and WiFi signal datasets.Different data dimensions were used for the data,and the length of the data was intercepted and processed.Then,Long Short-Term Memory(LSTM)and Convolutional Neural Networks were constructed to identify communication signals in ZigBee network.Simulation results show that LSTM has higher recognition accuracy than CNN,the recognition accuracy rate can reach 90.5%under low signal to noise ratio.