A compact model for speaker-adaptive training

TLDR

Speaker‑independent acoustic models suffer from variability caused by both phonetic differences and speaker differences, independent of speech content. The study proposes a new method for estimating continuous‑density HMM parameters for speaker‑independent speech recognition. The method decouples phonetic and speaker variation, jointly removes inter‑speaker variability while estimating SI HMM parameters, and is evaluated against standard SI training in supervised adaptation. The approach yields 19% and 25% word‑error‑rate reductions on 20K and 5K‑vocabulary tasks, indicating more efficient adaptation to test speakers.

Abstract

We formulate a novel approach to estimating the parameters of continuous density HMMs for speaker-independent (SI) continuous speech recognition. It is motivated by the fact that variability in SI acoustic models is attributed to both phonetic variation and variation among the speakers of the training population, that is independent of the information content of the speech signal. These two variation sources are decoupled and the proposed method jointly annihilates the inter-speaker variation and estimates the HMM parameters of the SI acoustic models. We compare the proposed training algorithm to the common SI training paradigm within the context of supervised adaptation. We show that the proposed acoustic models are more efficiently adapted to the test speakers, thus achieving significant overall word error rate reductions of 19% and 25% for 20K and 5K vocabulary tasks respectively.