Rapid speaker adaptation in eigenvoice space

TLDR

The paper proposes a model‑based speaker adaptation algorithm, the eigenvoice approach, to efficiently adapt acoustic models. It constrains the adapted model to a linear combination of a small set of orthogonal eigenvoice basis vectors derived from reference speakers, greatly reducing the number of parameters to estimate. Experiments on a small‑vocabulary letter‑recognition task show 16 % relative error reduction with one letter of adaptation data and 26 % with four letters, outperforming other adaptation methods.

Abstract

This paper describes a new model-based speaker adaptation algorithm called the eigenvoice approach. The approach constrains the adapted model to be a linear combination of a small number of basis vectors obtained offline from a set of reference speakers, and thus greatly reduces the number of free parameters to be estimated from adaptation data. These "eigenvoice" basis vectors are orthogonal to each other and guaranteed to represent the most important components of variation between the reference speakers. Experimental results for a small-vocabulary task (letter recognition) given in the paper show that the approach yields major improvements in performance for tiny amounts of adaptation data. For instance, we obtained 16% relative improvement in error rate with one letter of supervised adaptation data, and 26% relative improvement with four letters of supervised adaptation data. After a comparison of the eigenvoice approach with other speaker adaptation algorithms, the paper concludes with a discussion of future work.