Reassignment methods for robust time-frequency representations

In this paper, the reassignment method is extended to the robust local polynomial periodogram to process the nonstationary signals corrupted by impulse noise. Performance comparisons between robust spectrogram, the reassigned robust spectrogram, the robust local polynomial periodogram and the reassigned robust local polynomial periodogram (RrLPP) are presented. During computation of the local polynomial Fourier transforms, the length of overlap between the adjacent data segments is reduced to minimize the computational complexity. It shows that under the environment of impulse noise, with reduced overlap, the RrLPP and the reassigned robust local polynomial periodogram along frequency direction (RfrLPP) achieve much better signal concentrations than their counterparts without reassignment. It is also observed that the RfrLPP can achieve comparable results as the RrLPP, with reduced computational time.