Signal analysis by homomorphic prediction

Abstract

Two commonly used signal analysis techniques are linear prediction and homomorphic filtering. Each has particular advantages and limitations. This paper considers several ways of combining these methods to capitalize on the advantages of both. The resulting techniques, referred to collectively as homomorphic prediction, are potentially useful for pole-zero modeling and inverse filtering of mixed phase signals. Two of these techniques are illustrated by means of synthetic examples.