The Research Of Underwater Acoustic Pulse Signal Modulation Recognition Based On Time-Frequency Analysis

The main task for signal modulation recognition is to identify the signal modulation scheme through some kind of decision rules on the basis of signal detection and parameter estimation,providing the necessary information for subsequent tasks such as signal demodulation,information interferenceand signal monitor etc.The technology of signal modulation recognition arevaried,with the advance of digital image processing technology,applying the technologyof image processing to signal processingfieldprovides a new research direction for signal modulation recognition.As an effective method for non-stationary signal processing,time-frequency analysis candescribe the signal information from two dimensions including time domain and frequency domain.It could be observed that different pulse signals have obvious geometrydifferencesby analyzing signals’ time-frequency distribution images,thus the method of pulse signal modulation recognitionbased on the technology of image recognition is studied in the thesis.Firstly,thethesis introduces several kinds of common time-frequency analysis methods,including short-time Fourier transform,Wigner-Ville distribution and Cohen time-frequency distribution etc.Besides the common time-frequency analysis methods,the thesis studiesa kind oftime-frequency analysis method which based on the adaptive kernel,called adaptive kernel time-frequency distribution(AOK TFD).A series of image preprocessing methods are applied to enhance the time-frequency image and make the image features more obvious.Since the moment features can describe the shape of objects in the image,the thesis takes center distance,invariant moment and Pseudo-Zernike moment as feature parameters for image recognition.Finally,the BP neural network is applied to the pulse signal modulation recognition.The simulation resultsshow that the method of pulse signal modulation recognition by extracting moment features from time-frequency distribution imagesis feasible and effective.