Signal Detection And Recognition Method Based On Time-frequency Analysis Research

Under the background of modern communication, the signal received usingbroadband receiver is often a complex and fast changing signal. Subsequent signalprocessing faces enormous challenges very often because of signal itself often containthe complex modulation mode of spread spectrum, frequency hopping, time hopping etc.Compared to signal processing only in the time domain or frequency domain, thetime-frequency analysis technologies can intuitive access the distribution of the signalin time and frequency direction. So it can more efficient achieve of signal detection,parameter estimation, and modulation recognition.This dissertation based on the time-frequency analysis and studies its application insignal detection and identification. The work of the thesis is shown below:(1)The basic theory of time-frequency analysis is introduced. And then this thesishas a deep analysis of the STFT (Short Time Fourier Transform) and WVD(Wigner-Ville Distribution) transform features, advantages and disadvantages. Theeffect of diffusion algorithm and SPWVD (Smoothed pseudo WVD) time-frequencytransform in the cross term suppression is compared for the problem of how to eliminatethe WVD interference terms.(2) The application of common graphics processing algorithms in time-frequencyanalysis is studied. In order to access the accurately signal center frequency, arrival time,dwell time and signal level estimation, the graphic enhancements, graphic segmentation,corrosion, expansion and skeleton processing algorithm are used in this thesis. Aimingat the problem that the weak signal can hardly be detected, the processing method ofblock graphics is proposed and it effectively solves this issue.(3) Introduced the conception of multi-scales wavelet time-frequency analysis. Thedifferent effects of signal analysis with different wavelet scales are compared. For singlefrequency modulation signal, this thesis proposes a conception of optimal wavelet scale.The simulation finally shows the superiority of the optimal wavelet scale in signalprocessing.(4) When the signal under the influence of the shaping filter, the MFSK, BPSK, QPSK,8PSK,16QAM and64QAM signal modulation type recognition have beenachieved using the conception of wavelet scale histogram and the optimal wavelet scale.The characteristics of the signal code conversion time in the multi-scale wavelettransform are studied. And the signal symbol rate estimation is completed accurately inlow SNR.Finally, simulation results show the effectiveness of the method. The results showthat through the time-frequency analysis and combined with graphics processingalgorithms to achieve precisely the signal detection and parameter estimation. At lowSNR, the use of multi-scale wavelet transform can also get a good signal recognitioneffect and symbol rate estimation performance.