We present a condensed description and analysis of the joint submission for NIST SRE 2016, by Agnitio, BUT and CRIM (ABC). We concentrate on challenges that arose during development and we analyze the results obtained on the evaluation data and on our development sets. We show that testing on mismatched, non-English and short duration data introduced in NIST SRE 2016 is a difficult problem for current state-of-the-art systems. Testing on this data brought back the issue of score normalization and it also revealed that the bottleneck features (BN), which are superior when used for telephone English, are lacking in performance against the standard acoustic features like Mel Frequency Cepstral Coefficients (MFCCs). We offer ABC’s insights, findings and suggestions for building a robust system suitable for mismatched, non-English and relatively noisy data such as those in NIST SRE 2016.
Cite as: Plchot, O., Matějka, P., Silnova, A., Novotný, O., Sánchez, M.D., Rohdin, J., Glembek, O., Brümmer, N., Swart, A., Jorrín-Prieto, J., García, P., Buera, L., Kenny, P., Alam, J., Bhattacharya, G. (2017) Analysis and Description of ABC Submission to NIST SRE 2016. Proc. Interspeech 2017, 1348-1352, doi: 10.21437/Interspeech.2017-1498
@inproceedings{plchot17_interspeech,
author={Oldřich Plchot and Pavel Matějka and Anna Silnova and Ondřej Novotný and Mireia Diez Sánchez and Johan Rohdin and Ondřej Glembek and Niko Brümmer and Albert Swart and Jesús Jorrín-Prieto and Paola García and Luis Buera and Patrick Kenny and Jahangir Alam and Gautam Bhattacharya},
title={{Analysis and Description of ABC Submission to NIST SRE 2016}},
year=2017,
booktitle={Proc. Interspeech 2017},
pages={1348--1352},
doi={10.21437/Interspeech.2017-1498}
}