The ensemble speaker and speaking environment modeling (ESSEM) framework was
designed to provide online optimization for enhancing workable systems under real-world
conditions. In the ESSEM framework, ensemble models are built in the offline phase to
characterize specific environments based on local statistics prepared from those particular
conditions. In the online phase, a mapping function is computed based on the incoming
testing data to perform model adaptation. Previous studies utilized linear combination (LC) …