| More and more attention has been payed to interference processing technology because of the severe interference problems brought by the increasingly complex electromagnetic spectrum environment.As one of the current trends,machine learning is driving the development of communication anti-jamming technology to the direction of intelligence.Interference recognition is not only the prerequisite and foundation of anti-jamming,but also its key technology.The type of interference can be identified more efficiently and accurately by applying machine learning algorithms to interference recognition technology.After the interference signal is identified,the link adaptation technology is used to modify the parameters such as the transmission power,modulation and coding scheme and signal beam direction in real time according to different channel quality indicators,which can alleviate and suppress the interference to a certain extent.To complete the machine learning-based interference feature identification and classification jamming suppression,this paper combines the Support Vector Machine(SVM)algorithm and Convolutional Neural Network(CNN)in machine learning with the satellite communication system.In addition,the adaptive physical layer communication link of the wireless local area network is considered through SVM.The actual simulation verification is carried out on the Universal Software Radio Peripheral(USRP).Firstly,the current research status of machine learning in the field of interference detection and recognition is introduced.And at the same time,the paper summarizes the current research status of link adaptation related technologies.Besides,the research content and general arrangement of this paper are described briefly.Secondly,the principle of the SVM algorithm applied to the subsequent simulations of this paper is studied and discussed.And the flow chart of the algorithm in the field of feature extraction and classification is given.Then,aiming at the problem of image classification in this paper,the principle of CNN is outlined.And the convolutional neural network of this article is constructed with the relevant parameters.The satellite-ground communication system model is studied in chapter 3.And the low-orbit satellite downlink communication link is designed and built with reference to3GPP standards and proposals.The bit error rate(BER)and block error rate(BLER)performance under different maximum doppler shifts,different interference types,and variable interference strengths are compared in the simulations.By combining the SVM algorithm with the satellite-to-earth communication link,four eigenvalues of kurtosis,skewness,spectral flatness coefficient and frequency domain peak-to-average ratio are extracted from the power spectrum of the receiving signal.By using the four-dimensional combining features,the interference is classified and identified,and the classification accuracy of the four types of interference under different jamming-to-signal power ratio(JSR)are obtained by simulation.Then a convolutional neural network is constructed through the CNN image recognition algorithm to classify the spectrograms of the receiving signal under different interference types and different interference strengths.The classification accuracy under different JSRs are obtained,and the performance of the two algorithms is compared.The results show that the SVM algorithm can achieve an overall classification accuracy of 100%when the JSR reaches-1d B.Among them,the three types of interference,single-tone,multi-tone,and narrow-band interference are pecfectly classified when JSR reaches-16d B,-13d B,and-6d B,respectively.In the CNN classification results,the four types of interference of single-tone,multi-tone,sweep,and narrowband achieve the 100%classification accuracy when the JSR reaches-22d B,-18d B,-12d B,and-12d B,respectively.The classification performance of CNN is better than SVM.The interference adaptive link based on the IEEE 802.11a physical layer communication standard is built in chapter 4.The real-time switching of modulation and coding scheme under different interference types and strengths is realized by dividing the signal-to-noise ratio(SNR)interval and using the SNR estimated at the receiving end.Then,the type of interference is detected in real time by combining the SVM algorithm with the link model.Next,the N-sigma algorithm is applied to calculate the interference suppression threshold according to different interference types.The interference zeroing method is applied to the spectrum lines that exceed the threshold.In this way,we obtain the link transmission rate and bit error rate curve before and after interference suppression under different JSRs.The simulation results show that,the link transmission rate and BER performance of single-tone interference and multi-tone interference after interference suppression have been greatly improved compared with the results before interference suppression.When JSR exceeds-14d B,the transmission rate after interference suppression stabilizes at 54Mbps,and the bit error rate remains at around 8×10-5.After multi-tone interference suppression,the transmission rate is stable at 24Mbps,and the bit error rate is maintained at about 1×10-3when JSR is greater than-6d B.After sweep interference suppression,the transmission rate has been improved and the bit error rate is controlled to stablelize at about 6×10-3 within a certain JSR range.The bit error rate has increased with the increase of JSR,but it keeps being lower than the result before interference suppression.The suppression effect of narrow band interference has been the worst due to the wide interference spectrum.The zeroing of interference will lose part of the useful signal information.Interference suppression cannot increase the transmission rate,but it can reduce the bit error rate to 0.1 when it exceeds 0.1.Finally,a single-carrier communication software and hardware test platform and plan are designed.The communication link is built on the Matlab Simulink?platform,and the Pluto SDR equipped with the AD9363 RF transceiver chip is used as the RF transceiver.Additionally,the USRP is used to control GNU Radio to generate different types of interference.Then,a detailed theoretical analysis of coarse frequency estimation,symbol synchronization,carrier synchronization,phase offset compensation at the receiving end is carried out.Furthermore,the SVM algorithm is used to extract and classify the features of the power spectrum at the receiving end.And the classification accuracy curves under different JSRs are obtained by simulation.When the JSR reaches2d B,the four types of interference of single tone,multi-tone,sweep and narrowband are perfectly classified,which verifies the effectiveness of the classification algorithm used. |
PDF Full Text Request