The Research Of Trend Adaptive Decomposition Based On Compressive Sensing

Many of the signals in real life are multi-component signals whose frequency and amplitude are changing all the time.In the time-frequency analysis of multi-component signals,the precise calculation of the instantaneous frequency is very important,but the instantaneous frequency has no actual physical meaning because of the diversity and complexity of the oscillation modes.Therefore the effective decomposition method for multi-component signals is extremely important in the field of time-frequency analysis and has always been a hot issue in research.In the research,the concept of the trend line function of the signal was first proposed.Different from concepts of the intrinsic mode function and the product function,the trend line function was based on the characteristics of the signal itself.Secondly,an adaptive single-component decomposition method was proposed.Based on the compressed sensing theory and signal trend line function,the single-component decomposition of the signal was performed.First,the low-frequency trend component of the signal was extracted to maximize the original trend of the signal and reduce the high-frequency component pair.The effects of low frequency trends then achieve layer-by-layer decomposition of the signal from low to high frequencies.In the research,based on the adaptive single-component decomposition idea,the compressed sensing method was used to adaptively decompose the known signal.After the completed decomposition,the time-frequency analysis of the decomposed signal was performed,and the parameters such as the instantaneous frequency were obtained.The feasibility of this method was verified by simulation experiments with various types of trigonometric functions.The method was used in fault diagnosis,the number of components was significantly less than the traditional empirical mode decomposition,the power spectral density peak positioning was accurate,and the redundancy was reduced.In the application of human micro-motion feature extraction,the characteristic parameters calculation accuracy is better than the empirical mode and local mean decomposition,which proves that the method is feasible in practical applications.Figure 55;Table 9;Reference 60.