%Aigaion2 BibTeX export from Idiap Publications
%Saturday 21 December 2024 07:21:10 PM

@TECHREPORT{Prasad_Idiap-Internal-RR-54-2022,
         author = {Prasad, Amrutha and Khalil, Driss and Madikeri, Srikanth and Motlicek, Petr},
          title = {IDIAP SUBMISSION TO NIST LRE22 LANGUAGE RECOGNITION EVALUATION},
           type = {Idiap-RR},
         number = {Idiap-Internal-RR-54-2022},
           year = {2022},
    institution = {Idiap},
       abstract = {The Idiap submission to the NIST Language Recognition Evaluation (LRE) 2022 consists of three types of systems: (i) Random Forest (RF) and Support Vector Machine (SVM) classifiers trained on embeddings obtained from a pre-trained model from SpeechBrain, (ii) Kaldi-based x-vector-PLDA (Probabilistic Linear Discriminant Analysis) system trained with Kaldi, and (iii) Kaldi-based PLDA trained on the previously mentioned pre-trained model's embeddings. The score-level fusion (that is, linear combination of scores) of the RF and SVM classifiers in (i) was submitted as the primary system for the fixed condition. The score-level fusion of (ii) and (iii) were used as the alternative system. For the open condition, we used two Kaldi-based x-vector PLDA systems with score-level fusion, where additional data from the BABEL corpora was used to train the PLDA models. Our models were developed with Kaldi, PyTorch, SpeechBrain, and Scikit-learn toolkits.},
            pdf = {https://publications.idiap.ch/attachments/internals/2022/Prasad_Idiap-Internal-RR-54-2022.pdf}
}