## ABSTRACT

This chapter introduces the signal models Autoregressive (AR), moving average, and autoregressive moving average (ARMA). It explains power spectrum estimation using parametric methods such as MA model parameters, AR model parameters, and ARMA model parameters. The chapter deals with other spectrum evaluation methods like minimum variance method and eigenvalue algorithm method. It describes the cepstrum domain, and illustrates its use for pitch period measurement of speech signal. The chapter analyses higher order spectrum estimation called cumulant spectra. Stationary random signals are a class of signals of infinite duration and infinite energy. The average power is finite. Such signals are classified as power signals. Nonparametric methods are used to estimate the power spectrum. The Bartlett method or the method of averaging periodograms reduces the variance of the power density spectrum estimate. The Blackman–Tukey method uses the method of autocorrelation to find the power spectrum. It passes the autocorrelation via a smooth window to compensate error while calculating autocorrelation.