Quality Measures for Speaker Verification with Short Utterances
This work addresses the problem of reduced accuracy in speaker verification for short speech segments, which is incremental as it builds on existing system combination methods.
The paper tackles performance degradation in automatic speaker verification (ASV) systems with short utterances by incorporating quality measures of model parameters as supplementary information during system combination, resulting in considerable improvement in speaker recognition performance on NIST SRE corpora, especially for short durations.
The performances of the automatic speaker verification (ASV) systems degrade due to the reduction in the amount of speech used for enrollment and verification. Combining multiple systems based on different features and classifiers considerably reduces speaker verification error rate with short utterances. This work attempts to incorporate supplementary information during the system combination process. We use quality of the estimated model parameters as supplementary information. We introduce a class of novel quality measures formulated using the zero-order sufficient statistics used during the i-vector extraction process. We have used the proposed quality measures as side information for combining ASV systems based on Gaussian mixture model-universal background model (GMM-UBM) and i-vector. The proposed methods demonstrate considerable improvement in speaker recognition performance on NIST SRE corpora, especially in short duration conditions. We have also observed improvement over existing systems based on different duration-based quality measures.