English

Domain-Dependent Speaker Diarization for the Third DIHARD Challenge

Sound 2021-01-26 v1 Machine Learning Audio and Speech Processing

Abstract

This report presents the system developed by the ABSP Laboratory team for the third DIHARD speech diarization challenge. Our main contribution in this work is to develop a simple and efficient solution for acoustic domain dependent speech diarization. We explore speaker embeddings for \emph{acoustic domain identification} (ADI) task. Our study reveals that i-vector based method achieves considerably better performance than x-vector based approach in the third DIHARD challenge dataset. Next, we integrate the ADI module with the diarization framework. The performance substantially improved over that of the baseline when we optimized the thresholds for agglomerative hierarchical clustering and the parameters for dimensionality reduction during scoring for individual acoustic domains. We achieved a relative improvement of 9.63%9.63\% and 10.64%10.64\% in DER for core and full conditions, respectively, for Track 1 of the DIHARD III evaluation set.

Keywords

Cite

@article{arxiv.2101.09884,
  title  = {Domain-Dependent Speaker Diarization for the Third DIHARD Challenge},
  author = {A Kishore Kumar and Shefali Waldekar and Goutam Saha and Md Sahidullah},
  journal= {arXiv preprint arXiv:2101.09884},
  year   = {2021}
}

Comments

This work was presented in The Third DIHARD Speech Diarization Challenge Workshop

R2 v1 2026-06-23T22:28:41.311Z