Power Spectrum Identification for Adaptive Systems

This paper describes a method for identifying the parameters of a class of power spectra. In contrast to conventional methods of spectral analysis, the method assumes a particular form for the power spectrum and gives direct estimates of unknown parameters. Thus the method is faster than ordinary spectral analysis and can be easily implemented by a digital computer for use in an adaptive loop. The derivation of the estimates assumes that the power spectrum has no zeros, and is based on well-known results in the theory of autoregressive schemes. Some ways of extending the results to the case where zeros are present in the spectrum are suggested. The method can also be used as a prewhitening technique in conjunction with ordinary spectral analysis.