Anthony Larcher

Phonetically-constrained PLDA modeling for text-dependent speaker verification with multiple short utterances

The importance of phonetic variability for short duration speaker verification is widely acknowledged. This paper assesses the performance of Probabilistic Linear Discriminant Analysis (PLDA) and i-vector normalization for a text-dependent verification task. We show that using a class definition based on both speaker and phonetic content significantly improves the performance of a state-of-the-art system. We also compare four models for computing the verification scores using multiple enrollment utterances and show that using PLDA intrinsic scoring obtains the best performance in this context. This study suggests that such scoring regime remains to be optimized.

I-vectors in the context of phonetically-constrained short utterances for speaker verification

Short speech duration remains a critical factor of performance degradation when deploying a speaker verification system. To overcome this difficulty, a large number of commercial applications impose the use of fixed pass-phrases. In this context, we show that the performance of the popular i-vector approach can be greatly improved by taking advantage of the phonetic information that they convey. Moreover, as i-vectors require a conditioning process to reach high accuracy, we show that further improvements are possible by taking advantage of this phonetic information within the normalisation process. We compare two methods, Within Class Covariance Normalization (WCCN) and Eigen Factor Radial (EFR), both relying on parameters estimated on the same development data. Our study suggests that WCCN is more robust to data mismatch but less efficient than EFR when the development data has a better match with the test data.

Modelling the alternative hypothesis for text-dependent speaker verification

This paper describes text-dependent speaker verification as a task involving four classes of trials depending on whether the target speaker or an impostor pronounces the expected pass-phrase or not. These four classes are used to reformulate the log-likelihood ratio traditionally used in text-independent speaker verification. Three formulations of the alternative hypothesis are considered, leading to three new expressions of the verification score. Experiments performed on the publicly available RSR2015 database show a significant improvement compared to existing baseline scores. A relative gain up to 61% in term of minimum cost is achieved when considering that the alternative hypothesis is the union of three sub-hypotheses corresponding to the three existing classes of impostures.

Speaker verification performance with constrained durations

Over the last decade speaker recognition has witnessed significant advances, with successful developments in Factor Analysis (FA) and more recently i-vectors, more than halving the error rates achieved by the classical UBM/GMM approach. However when very short duration utterances are considered, it is known that these improvements are much less. This paper begins with a review of the recent developments of i-vector systems with a focus on short test duration, in the region of 10 seconds or less. Experimental results are then presented showing that error rates rise from approximately 5% to 18% when the test duration is systematically reduced from 30 seconds to just 3 seconds. Interestingly, with the 30 seconds condition the i-vector error rate is in the region of half that of the corresponding UBM/GMM system. Nevertheless, when the test segments are just 3 seconds duration then the error rates of the 2 systems systems are very similar. All experiments relate to the short-short condition of the NIST 2008 SRE, but with the test segment duration systematically reduced.