Matching and fusing signal-estimation errors for similarity-based pattern classification

Abstract

Error estimation using different optimal models for signal processing has been an active research field in data analysis such as speech recognition, image analysis, geophysics, and earth science. A popular direction of research in pattern classification is to develop computational models for comparing objects being either abstract or physical based on some measure of similarity or dissimilarity. This paper explores some linear-prediction models for deriving signal estimation errors and their fusion for similarity-based pattern classification.