Speaker Clustered Regression-Class Trees for MLLR Adaptation

Citation

Mandal, A., Ostendorf, M., & Stolcke, A. (2006). Speaker clustered regression-class trees for MLLR adaptation. In INTERSPEECH.

Abstract

A speaker clustering algorithm is presented that is based on an eigenspace representation of Maximum Likelihood Linear Regression (MLLR) transformations and is used for training cluster-dependent regression-class trees for MLLR adaptation. It is shown that significant automatic speech recognition (ASR) system performance gains are possible by choosing the best regression-class tree structure for individual speakers. To take advantage of the potential gains, an algorithm for combining the MLLR mean transformations from cluster-specific trees is described that effectively results in a soft regression-class tree. In conversational speech recognition, only small overall improvements are obtained, but the number of speakers that have performance degradation due to adaptation is reduced by over 70%.

Index Terms: speech recognition, speaker adaptation, speaker clustering, regression class trees.

↓ Download

Read more from SRI

  • surgeons around a surgical robot

    The SRI research behind today’s surgical robotics

    Intuitive’s da Vinci 5 system represents a major leap in robotic-assisted medicine. It all started at SRI, which continues to advance teleoperation technologies.

  • a collage of digital graphs

    A banner year for quantum

    SRI-managed QED-C’s annual report on quantum trends captures an industry accelerating rapidly from technical promise toward major global impact.

  • ICE Cube containing SRI’s aerogel experiment, photographed prior to launch. Source: Aerospace Applications North America

    An SRI carbon capture experiment launches into space

    By synthesizing carbon-absorbing aerogels in microgravity, SRI research will give us a rare glimpse into how these materials could be radically improved.