Title :
Spectral estimation using an improved comb filter method
Author_Institution :
Dept. of Electr. & Electron. Eng., North Dakota State Univ., Fargo, ND, USA
Abstract :
The comb filter is used with some improvement in energy compensation, segmenting-overlapping-averaging, and lag weighting with symmetrization, so that the resultant spectrum is closer to the true spectrum than that when using the general comb filter method. The measurement of power spectrum in a short duration of the features of a designed bandpass filter by using a white noise excitation. Two important parameters, D and R80, are used to compare the results derived from power spectral methods. D is the deviation of the estimated spectrum from the true spectrum, and R80 is the spectrum range in which 80% points fall. Some of numerical values of these two parameters obtained from experiment are listed. It appears from comparison that the improved comb filter method is more accurate than the general comb filter method. A plot of power spectrum estimated by the improved comb filter shows that it is a quite satisfactory approach to power spectral estimation
Keywords :
filtering and prediction theory; parameter estimation; spectral analysis; bandpass filter; comb filter; energy compensation; lag weighting; power spectrum; segmenting-overlapping-averaging; short duration; white noise excitation; Band pass filters; Energy loss; Neodymium; Noise measurement; Power engineering and energy; Power measurement; Signal processing; Sonar; Teeth; White noise;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 1991. ICASSP-91., 1991 International Conference on
Conference_Location :
Toronto, Ont.
Print_ISBN :
0-7803-0003-3
DOI :
10.1109/ICASSP.1991.150122