The Interference Detection Of Wireless Communication Based On Deep Learning

The limited spectrum resources and the explosive growth of communication equipment have made spectrum sensing one of the key technologies to improve the efficiency of spectrum utilization.In deep spectrum sensing,it is often necessary to detect and recognize unlicensed or interfering signals.The traditional interference detection method is mainly realized by analyzing the features such as the spectrum and power of the received signal.But for some interference signals that have the same frequency as the original signals and are generated simultaneously,it is difficult to detect and extract using traditional feature engineering methods.This paper will study this kind of special time-frequency overlapped interference signal,propose a wireless communication interference detection algorithm based on deep learning,and separate the interference signal and noise to support the analysis and recognition of the interference signal.The main contents of this paper include the following three aspects:First,an original signal prediction model based on deep learning is established,and the error between the predicted signal and the original signal is used as a feature for interference detection.Analyze and compare the prediction ability of various neural network structures on the signal,and improve the signal prediction performance under the condition of low signal-to-noise ratio by combining the training method with noiseless samples.After selecting the model with the best prediction performance to calculate the error feature,use this feature as an input to the SVDD model for interference detection.This method can achieve a certain effect when the SNR is high.It can not only detect the presence of the interference signal,but also determine the specific position of the interference signal in the time-domain waveform of the received signal.Then,in order to solve the problem of poor interference detection under the condition of low SNR,the idea of using signal reconstruction instead of signal prediction was proposed.An original signal reconstruction model based on AutoEncoder was built and the error between the original signal and the reconstructed signal was calculated as the feature to do interference detection.The final decision is made based on the SVDD algorithm,too.Experiments show that the difficulty of signal reconstruction by neural networks is lower than that of signal prediction,which makes the reconstructed signal very close to the original transmitted signal,so the interference detection performance using this algorithm is greatly improved.Finally,the deep neural network based on the reconstructed signal can separate the interference signal and noise from the original transmitted signal.The paper analyzes the possibility of further recognition with the separated interference signal,and performs the research of modulation recognition and radio frequency fingerprinting recognition on the separated interference signal.The experiment proves that the separated interference signal preserves most of the features of the original interference signal and achieves ideal results in both modulation recognition and radio frequency fingerprinting recognition.