Detection Based On The Morphology Of Time-Frequency Distribution

As one of the efficient tools to analyze and process nonstationary signals, the time-frequency analysis has been widely applied in detecting nonstationary signals from noises. The modulus of the time-frequency distribution of a signal can be regarded as a grayscale image, namely, the time-frequency image of the signal. In the time-frequency image of a signal contains the time-frequency information and these information is conveyed by different geometrical objects in the image.Based on time-frequency concentration of signals and random dispersion of noise in time-frequency images, we propose a new method to detect signal from noise using the morphological filtering of time-frequency images. The detection method is realized in three steps. First, the received signal in time domain is transferred to time-frequency domain by the time-frequency analysis and the time-frequency images are obtained. Secondly, the thresholding process is applied to time-frequency images to transfer them to binary images. Thirdly, we use the morphological filtering to extract relevant geometric objects in the images and then the total energy over these extracted objects is used to decide whether a signal is present or absent. The simulation results show that the proposed method can effectively detect signal from noise when the waveform of the signal is unknown.