/sre2016

code used in the NIST SRE2016 challenge.

Primary LanguageC++

Source code used in the NIST SRE2016 challenge

In addition to the main bash script (run.sh), this repository contains some code augmentations specific to Kaldi:

NOTE: NDA Hasn't improved LDA results in our in-home experiments. May contain some bugs in the implementation. Feel free to use this project in any capacity on top of your kaldi-based SID systems. Please let me know if you have any comments, or corrections.

Dependencies:

  • Kaldi
  • GLPK (only required for i-vector clustering)

Citations:

Shokouhi, N., & Hansen, J. H. (2015). Probabilistic linear discriminant analysis for robust speaker identification in co-channel speech. In Sixteenth Annual Conference of the International Speech Communication Association.
Zhang, C., Bahmaninezhad, F., Ranjan, S., Yu, C., Shokouhi, N., & Hansen, J. H. (2016). UTD-CRSS systems for 2016 nist speaker recognition evaluation. arXiv preprint arXiv:1610.07651.

Navid Shokouhi
2016