Signal processing using the Teager Energy Operator and other nonlinear operators

Abstract

The Teager Energy Operator is a nonlinear operator defined for both continuous and discrete time signals. In this thesis we define different versions and extensions of the one-dimensional operator, and then look at the properties and positivity of the operator. Its connection to Volterra systems is explained. The Teager Energy Operator is used for demodulation of AM-FM signals, and we look at how this can be done using the operator and nonlinear methods (Hilbert transform, Prony's, Modified covariance, etc). As the Teager Energy Operator is sensitive to noise, the multiband demodulation technique is introduced to refine the estimates. The performance of linear and nonlinear differentiators is investigated. An application to speech analysis is shown, using a multiband method to locate formants in speech. Finally, a performance comparison in different levels of noise is done.