Improved multi-scale chirplet sparse signal decomposition method based on Fractional Fourier Transform

In view of the large computation of the global searching algorithm of basic function,an improved multi-scale chirplet sparse signal decomposition method(IMSCSD) is proposed in which the basic function is ascertained by Fractional Fourier Transform(FRFT).The symmetric items of real signal's FRFT are removed by the FRFT of the analytic signal,and the FRFT amplitude spectrum is searched by two-step peak searching method to ascertain the best basic function.The efficiency and anti-noise performance of the IMSCSD is checked up,and IMSCSD is applied to decompose the non-stationary signal with multi-components whose frequencies change curvedly.The results show that the symmetry items of the FRFT of real signal is removed effectively by the FRFT of corresponding analytic signal;the basic function is ascertained by FRFT rapidly,precisely and robustly without the help of experience,and the method has no relation to frequency offset and frequency changing rate,so the large computation of global searching algorithm is overcome effectively;the signal with multi-components whose frequencies change curvedly is decomposed fast and precisely by IMSCSD,so the IMSCSD is very suitable for the evaluation of instantaneous frequencies and separation of targeted component of the non-stationary signal of rotational machine under transient condition.