Model-Based Approaches to Handling Uncertainty

摘要: A powerful approach for handling uncertainty in observations is to modify the statistical model of data appropriately reflect this uncertainty. For task noise-robust speech recognition, requires modifying an underlying “clean” acoustic be representative a particular target environment. This chapter describes concepts model-based noise compensation robust recognition and how it can applied standard systems. The will then consider important practical issues. These include i) environment parameter estimation; ii) efficient likelihood calculation; iii) adaptive training handle multi-style data. conclude by discussing limitations current approaches research options address them.