Title :
Order recursive parametric bispectrum estimation
Author_Institution :
Dept. of Electr. Eng., Wright State Univ., Dayton, OH, USA
Abstract :
Order-recursive computation of autoregressive (AR) parameters from cumulants is addressed. If the cumulant matrix is neither Toeplitz nor symmetric, it is shown that using the Frolenius-Schur block matrix inversion formula the inverse of the p-dimensional cumulant matrix can be updated from the (p-1)-dimensional inverse with O(p2) operations. When compared to batch mode computation, the proposed algorithm reduces the computational requirement for order-recursive calculation of the AR parameters. When the cumulant matrix is also nonsymmetric Toeplitz, further reduction in computation is obtained using Trench´s algorithm
Keywords :
matrix algebra; parameter estimation; recursive functions; spectral analysis; statistical analysis; AR parameters; Frolenius-Schur block matrix inversion formula; Trench algorithm; autoregressive parameters; cumulant matrix; nonsymmetric Toeplitz matrix; order recursive parametric bispectrum estimation; Autocorrelation; Costs; Digital signal processing; Equations; Higher order statistics; Recursive estimation; Signal processing algorithms; Spectral analysis; Symmetric matrices;
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.1991.150209