Domain adaptation based Speaker Recognition on Short Utterances

TL;DR

This study investigates how domain adaptation techniques affect speaker recognition performance on short utterances, highlighting the diminishing benefits of in-domain models and proposing a modified IDV compensation method to mitigate domain mismatch.

Contribution

The paper introduces a novel modified IDV compensation method that improves out-domain PLDA performance, especially for short utterances, by addressing dataset mismatch.

Findings

In-domain PLDA outperforms out-domain PLDA on full-length utterances with over 28% improvement.

Performance gains from IDV compensation decrease as utterance length shortens due to phonetic variability.

IDV compensation yields 26% and 14% improvements on SWB and NIST datasets, respectively.

Abstract

This paper explores how the in- and out-domain probabilistic linear discriminant analysis (PLDA) speaker verification behave when enrolment and verification lengths are reduced. Experiment studies have found that when full-length utterance is used for evaluation, in-domain PLDA approach shows more than 28% improvement in EER and DCF values over out-domain PLDA approach and when short utterances are used for evaluation, the performance gain of in-domain speaker verification reduces at an increasing rate. Novel modified inter dataset variability (IDV) compensation is used to compensate the mismatch between in- and out-domain data and IDV-compensated out-domain PLDA shows respectively 26% and 14% improvement over out-domain PLDA speaker verification when SWB and NIST data are respectively used for S normalization. When the evaluation utterance length is reduced, the performance gain by IDV also reduces as short utterance evaluation data i-vectors have more variations due to phonetic variations when compared to the dataset mismatch between in- and out-domain data.