A novel method of model order selection and separation of spectral structures is proposed for practical signals which are composed of multiple structured spectra. The input time sequence signal is considered to have multiple spectral structure, i.e., a main spectral structure and a residual spectral structure. The main structure is determined by the signal dominant power spectral component of the input signal and the residual structure is defined by the residual power spectral component after the dominant power spectral component is removed from the whole spectral structure. The method first estimates the dominant spectrum of the main structure using the AR (autoregressive) model with order p[AR (p)] and then estimates the spectrum of the residual structure using the AR model with order q[AR (q)]. By computer simulation, the method is proved to give a good solution to the problem of reliable spectral estimation.
|ジャーナル||ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings|
|出版ステータス||Published - 1990|
|イベント||1990 International Conference on Acoustics, Speech, and Signal Processing: Speech Processing 2, VLSI, Audio and Electroacoustics Part 2 (of 5) - Albuquerque, New Mexico, USA|
継続期間: 1990 4月 3 → 1990 4月 6
ASJC Scopus subject areas