Class 2 homomorphic estimation and detection of multiplicatively-combined signals

Abstract

The authors discuss separating and detecting signals which are combined multiplicatively where the system input is the product of the signal and the noise or the unwanted component. This system can be applied to amplitude modulation, fading channels, audio dynamic range compression and expansion, automatic gain control, radar, sonar, image enhancement, and music applications. When either the model for the signal expectation spectrum or that for the noise expectation spectrum, but not both, is known a priori, the Class 2 algorithm has to be used. The unknown spectral model must be explicitly estimated a posteriori from the ideal filter input. Optimum estimation and detection filters are developed by using homomorphic techniques to separate these signals.<>