DSP-based AR spectral estimation with zoom effect for NMR spectroscopy

Autoregressive (AR) spectral estimation is a high-resolution method based on the modeling of time series by an AR model. AR parameters are estimated by a fast recursive least squares (FRLS) algorithm with a transversal structure, the fast Kalman (FK) algorithm. To implement it on a DSP96002 processor the algorithm is stabilized by using the method of Benallal that avoids the accumulation of round-off errors. For particular applications, only a part of the spectrum is of interest. For that case, a zoom function is described that can be used to yield local spectra with better resolution. For this purpose, the complex form of the stabilized FK (SFK) is used. An application of this AR spectral analysis to nuclear magnetic resonance (NMR) spectroscopy is shown.<
>