Fast Signal Detection And Specific Signal Recognition Based On Statistical Learning

In non-cooperative mode,if you want to identify some signals that carry a specific frame header,you generally need to estimate all the parameters of the signal.Due to the influence of factors such as channel environment and receiver performance,the shape of the spectrum obtained by the reconnaissance receiver for multiple scans of the same signal and the same frequency band is slightly different.Causing difficulties in cognizing signals and spectrum,so it is of great practical significance to study fast signal detection and specific signal recognition in a non-cooperative environment.Based on the above problems,this article focuses on the rapid detection of signal spectrum and the identification of specific signals.Using statistical learning methods to analyze a large amount of spectrum data in detail,so as to avoid the influence of channel environment and receiver performance on signal detection and estimation,and obtain more accurate results of spectrum center frequency and bandwidth estimation as a basis for identifying specific signals.The main work of this article is as follows:1.A fast signal detection method based on statistical learning is designed,which can detect all multiple signals that meet the detection requirements,on this basis,this paper proposes an improved grid-based K-means clustering algorithm.By reducing the number of repartitioned signal points in the iterative solution to achieve the purpose of speeding up the calculation.Simulation results prove that: compared with the original K-means clustering algorithm,the detection speed of the improved method is increased almost 6.1 times,while the detection accuracy rate remains high,and the false alarm rate / missing alarm rate of detection is low,which can meet performance requirements for rapid detection of specific signals.2.Two hybrid identification methods for multiple specific signals based on frame headers are proposed.The first method uses a specific frame header set by the user to generate a time-domain reference waveform template that correlates with the baseband waveform after down-conversion of the received signal to detect correlation peak;the second method generates a frequency-domain reference waveform template with a specific frame header,and performs correlation matching with the received signal shorttime Fourier transform result to obtain the correlation between the two,and transfers the recognition process to the relevant subspace.The supervised K-means algorithm completes the statistical analysis of the correlation and completes the identification of specific signals.The two schemes are simple in structure and easy to implement.The recognition accuracy of each modulated signal under different signal-to-noise ratios has been simulated and verified,which proves that this scheme can effectively complete the rapid recognition of specific signals.3.A client-side signal fast detection and specific signal recognition software based on the QT development platform is designed and implemented,combined with relevant theory and actual engineering application requirements,realized convenient parameter configuration,perfect interface display,real-time local storage and file management,and fast signal functions such as detection and identification of specific signals and rapid function expansion.After the actual performance test of the software,it has proved that the software has good detection and recognition performance.