According to the fact that there exists illusive components when noisy signal is under Hilbert-Huang transform

an improved singular value decomposition (SVD) was firstly proposed for two improvements: one was that the reconstructed phase space was employed instead of traditional matrix such as Hankle matrix in order to remove redundancy

the other was that a singular value energy entropy component difference method was proposed

making it easier to determine order of reconstruction singular value. Secondly

a spectrum ratio method was raised to identify illusive components

making it easier to identify illusive components. Intrisinc mode function (IMF) was obtained by empirical mode decomposition (EMD) and trend term was detrended

then the signal was reconstructed for subsequent SVD operation; the denoised signal was obtained and illusive component was eliminated

lastly the signal time-frequency distribution was calculated. The SNR raised 5.5% when the jointly proposed method employed in simulation signal

the illusive component identity ratio raised to 100%

two-span rotor fault vibration signal analysis results show that the jointly proposed method is effective in time-frequency distribution calculation and identification of noisy signal illusive components.