Composite modeling of nonstationary signals

✍ Scribed by M. Boudaoud; L.F. Chaparro

• Publisher: Elsevier Science
• Year: 1987
• Tongue: English
• Weight: 633 KB
• Volume: 324
• Category: Article
• ISSN: 0016-0032
• DOI: 10.1016/0016-0032(87)90377-2

✦ Synopsis

Most temporal signals of practical interest are nonstationary and need to be modeled using time-varying systems. In this paper a composite model for these signals is proposed which accounts for nonstationarities in the mean and the autocorrelation functions. Unbiased and consistent time-varying estimators for the mean and the variance functions are studied and used to produce zero-mean, constant-variance signals that can be modeled using autoregressive systems with time-varying coeficients. The identification of the coejicients is implemented recursively using the parameterization of the coeficients by a set of basis functions. We illustrate the application of the composite model in the analysis and synthesis of speech and in the estimation of instantaneous frequencies in radar return si~qnals.