Quality Measures for Speaker Verification with Short Utterances

Arnab Poddar 1 Md Sahidullah 2 Goutam Saha 1
2 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication
Inria Nancy - Grand Est, LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
Abstract : The performances of the automatic speaker verification (ASV) systems degrade due to the reduction in 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 a 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. Considerable improvement is found in performance metrics by the proposed system on NIST SRE corpora in short duration conditions. We have observed improvement over state-of-the-art i-vector system.
Complete list of metadatas

Cited literature [67 references]  Display  Hide  Download

https://hal.inria.fr/hal-01998376
Contributor : Md Sahidullah <>
Submitted on : Tuesday, January 29, 2019 - 3:44:35 PM
Last modification on : Monday, January 6, 2020 - 5:36:05 AM
Long-term archiving on: Tuesday, April 30, 2019 - 4:40:47 PM

File

DSP_ASV.pdf
Files produced by the author(s)

Identifiers

Collections

Citation

Arnab Poddar, Md Sahidullah, Goutam Saha. Quality Measures for Speaker Verification with Short Utterances. Digital Signal Processing, Elsevier, 2019, 88, pp.66-79. ⟨10.1016/j.dsp.2019.01.023⟩. ⟨hal-01998376⟩

Share

Metrics

Record views

218

Files downloads

315