Multipath Interference Suppression Method Based On Convolutional Newral Network In Chaos Baseband Wireless Communication Systems

In a complex wireless communication channel,inter-symbol interference(ISI)caused by signal multipath propagation seriously affects communication performance.How to mitigate inter-symbol interference in chaotic baseband wireless communication systems is an urgent problem to be solved.Studies have shown that inter-symbol interference in chaotic baseband wireless communication systems can be reduced by calculating suboptimal decoding thresholds.However,there is a problem in the method of calculating the sub-optimal decoding threshold:since no future symbol information is used,only partial inter-symbol interference(inter-symbol interference caused by the historical symbol information)can be eliminated.To solve this problem,the main work of this article is as follows:(1)An inter-symbol interference prediction method based on CNN to predict future symbols is proposed.This method combines a convolutional neural network with a deep learning structure to predict future symbol information that has not been received based on historical symbol information that has been received,thereby eliminating inter-symbol interference caused by future symbol information,and calculating suitable for time-varying decoding threshold to solve the problem that the inter-symbol interference caused by future symbols cannnot be eliminated in the suboptimal threshold method.By simulation and experiment test in real outdoor scenes,the different decoding methods are used to decode the received chaotic signals.The simulation and experiment results show that,compared with the traditional decoding method and sub-optimal decoding threshold method,the proposed method has the lowest bit error rate under different signal-to-noise ratio conditions,which proves the effectiveness of the proposed method(2)A chaotic baseband wireless communication method based on CNN direct decoding is proposed.This method takes advantage of the superior performance of the convolutional neural network in classification problems.By proper training the convolutional neural network,CNN can directly output corresponding symbol value according to the input waveform information.Compared with the existing methods,the proposed method can eliminate the threshold calculation and decoding process according to the threshold.By comparing the simulation and experiment results of the proposed method with the traditional method and the suboptimal decoding threshold,we learn that the proposed method achieves the lowest BER under different signal-to-noise ratio.