Spread Spectrum Signals Recognition Scheme Based On Multi-dimensional Parameter Extraction And SVM

Spread spectrum communication is widely used in various fields of modern communication because of its low probability of interception,high anti-interference ability,safety and concealment.In the field of jamming countermeasure,with the continuous development of radio reconnaissance technology,to achieve safe and reliable communication,the opposing sides will use spread spectrum technology for communication.At the same time,they will also try to detect and intercept the enemy's communication signals.Due to the widespread use of spread spectrum signals in the field of jamming countermeasure,the recognition of spread spectrum signals is particularly important.The current spread spectrum signals recognition schemes can identify few signal types,and when the received signal contains multiple types of spread spectrum signals,the accuracy of the recognition scheme will decrease.Therefore,this thesis studies the automatic recognition scheme of spread spectrum signals and interference signals in the field of jamming countermeasure,including frequency hopping signal,direct sequence spread spectrum signal,hybrid spread spectrum signal,single tone signal,multitone signal and rectangular pulse signal.Selecting appropriate classification parameters is a prerequisite for efficient signal recognition and classification.Therefore,this thesis first analyzes the characteristics of various signals in the three dimensions of time domain,frequency domain,and time-frequency domain,extracts six classification parameters to characterize the differences between various types of signals,including autocorrelation peak-to-average ratio,autocorrelation second order moment,spectrum bandwidth,power spectrum cancellation gain,time-frequency matrix normalized column mode and normalized row mode.The value distribution of various signals on these six classification parameters is analyzed through simulation,which proves the rationality of the selection of these classification parameters.On the basis of these six classification parameters,this thesis combines the multi-class support vector machine algorithm to design a spread spectrum signals recognition scheme based on multi-dimensional parameter extraction and SVM.The recognition scheme is used in non-cooperative receiving scenarios,and consists of two modules: offline training and online recognition.The offline training module extracts multi-dimensional parameters of various training signals,then construct a training set,and obtains a set of decision functions through training,thus providing support for the online recognition module.The online recognition module first extracts the multi-dimensional parameters of the received signal to obtain the signal feature vector,and uses the decision function set to determine the feature vector according to the voting strategy,thereby realizing the recognition and classification of the received signal.After that,the scheme proposed in this thesis and the recognition scheme based on decision tree are simulated,results show that the accuracy rate of the proposed scheme in higher.Through simulation,the minimum signal number required by the recognition scheme in this thesis is obtained on the premise of ensuring the recognition accuracy.Finally,the accuracy rate of this scheme under Gaussian white noise channel and multipath Rayleigh fading channel is analyzed,results show that the recognition scheme of this thesis can achieve a higher accuracy rate under different channel conditions.