Multiple-window cepstrum analysis for estimation of periodicity
Publikation/Tidskrift/Serie: IEEE Transactions on Signal Processing
Dokumenttyp: Artikel i tidskrift
Förlag: IEEE--Institute of Electrical and Electronics Engineers Inc.
Multiple windows have been used in spectrum analysis to reduce variance. In this paper, a multiple-window approach is suggested in cepstrum analysis, where the period of the system response should be estimated. The logarithm of the spectrum as well as the liftering are approximated by a division of two multiple-window spectra. The method is useful especially in the case of short periods of the periodic signal and large spectral dynamics of the other signal, i.e., the case with severe overlap in the cepstrum. Examples are given for estimation of periods of surface waves traveling around the circumference of tree trunks.
- Probability Theory and Statistics
- real-valued cepstrum
- homomorphic deconvolution
- multiple-window spectrum analysis
- Statistical Signal Processing Group
- ISSN: 1053-587X