AR_SPA decomposes an AR-spectrum into its compontents
[w,A,B,R,P,F,ip] = ar_spa(AR,fs,E);

INPUT:
AR   autoregressive parameters
fs    sampling rate, provide w and B in [Hz], if not given the result is in radians
E     noise level (mean square),  gives A and F in units of E, if not given as relative amplitude

OUTPUT
w	center frequency
A     Amplitude
B     bandwidth
- less important output parameters -
R	residual
P	poles
ip	number of complex conjugate poles
real(F)     	power, absolute values are obtained by multiplying with noise variance E(p+1)
imag(F)	assymetry, - " -

All input and output parameters are organized in rows, one row
corresponds to the parameters of one channel

see also ACOVF ACORF DURLEV IDURLEV PARCOR YUWA

REFERENCES:
[1] Zetterberg L.H. (1969) Estimation of parameter for linear difference equation with application to EEG analysis. Math. Biosci., 5, 227-275.
[2] Isaksson A. and Wennberg, A. (1975) Visual evaluation and computer analysis of the EEG - A comparison. Electroenceph. clin. Neurophysiol., 38: 79-86.
[3] G. Florian and G. Pfurtscheller (1994) Autoregressive model based spectral analysis with application to EEG. IIG - Report Series, University of Technolgy Graz, Austria.