Font Size: a A A

Improving The Performance Of Time-frequency Distribution By Adopting The Thought And Method In Image Processing

Posted on:2015-02-03Degree:MasterType:Thesis
Country:ChinaCandidate:A RenFull Text:PDF
GTID:2308330464970242Subject:Measuring and Testing Technology and Instruments
Abstract/Summary:PDF Full Text Request
Along with the continuous development in science, the signals collected from the real word are becoming more and more complicated, the traditional Fourier transform can no longer meets the needs of signal analysis. People want to understand the specific situation of the frequency change over time, so the joint time-frequency distribution emerges at the right moment. With the efforts of the researchers, a variety of time-frequency analysis methods are proposed in succession, they apply to different signals, these methods have been applied to telecommunication, military, medical, fault detection and other fields, then also put forward a lot of standards and methods to evaluate the performance of the time-frequency distribution.Among all the time-frequency analysis methods, the(pseudo) Wigner-Ville distribution in quadratic time-frequency transformation has better energy concentration. Yet in dealing with double-component or multi-component signals, they are disturbed by cross-terms. It is the efforts of suppressing cross-terms that promote the development of time-frequency analysis: S-method, multiwindow S-method and some other methods adopt frequency window to realize the suppression of cross-terms, but reduces the concentration of energy; exerting Hough transformation on WVD(PWVD), the auto-terms and cross-terms present different aggregation properties in the parameter space, so it can achieve the goal; designing best kernel function, etc.Most of these methods are the compromise between the suppression of cross terms and the reservation of auto-terms. This paper introduces the idea of image segmentation to(pseudo) Winger-Ville distribution, when the auto-terms are separated, it can cut out the cross-terms from the Wigner-Ville distribution, and would not do harm to the auto-terms. Achieve the goal of completely eliminating cross-terms, acquire a high concentration of energy.The other problem in time-frequency analysis is for double-component signals that amplitude of all components are in the same, after added noise, the accuracy of instantaneous frequency estimation will be greatly reduced. Inspired by the way of dealing with the noise in image processing, this article adopts the method of imageenhancement to process time-frequency distribution matrix, using kernel functions to smooth it, the noisy points are restrained effectively, as the experimental data shows, the accuracy of instantaneous frequency estimation is greatly improved.In this paper, the time-frequency analysis and image processing are combined effectively, it belongs to the interdisciplinary research. The relevant experimental data prove the feasibility and the research significance of the proposed method.
Keywords/Search Tags:Time-Frequency Analysis, (Pseudo) Wigner-Ville Distribution, Cross-Terms, Instantaneous Frequency, Image Segmentation, Image Enhancement
PDF Full Text Request
Related items