Publication Details
Training Speaker Embedding Extractors Using Multi-Speaker Audio with Unknown Speaker Boundaries
Mošner Ladislav, Ing. (DCGM)
Plchot Oldřich, Ing., Ph.D. (DCGM)
Rohdin Johan Andréas, M.Sc., Ph.D. (DCGM)
Silnova Anna, M.Sc., Ph.D. (DCGM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Černocký Jan, prof. Dr. Ing. (DCGM)
Speaker Embedding Extractors, Multi-Speaker Audio, Unknown Speaker Boundaries
In this paper, we demonstrate a method for training speaker em- bedding extractors using weak annotation. More specifically, we are using the full VoxCeleb recordings and the name of the celebrities appearing on each video without knowledge of the time intervals the celebrities appear in the video. We show that by combining a baseline speaker diarization algorithm that re- quires no training or parameter tuning, a modified loss with aggregation over segments, and a two-stage training approach, we are able to train a competitive ResNet-based embedding extractor. Finally, we experiment with two different aggregation functions and analyze their behaviour in terms of their gradients.
@inproceedings{BUT179781,
author="Themos {Stafylakis} and Ladislav {Mošner} and Oldřich {Plchot} and Johan Andréas {Rohdin} and Anna {Silnova} and Lukáš {Burget} and Jan {Černocký}",
title="Training Speaker Embedding Extractors Using Multi-Speaker Audio with Unknown Speaker Boundaries",
booktitle="Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH",
year="2022",
journal="Proceedings of Interspeech",
volume="2022",
number="9",
pages="605--609",
publisher="International Speech Communication Association",
address="Incheon",
doi="10.21437/Interspeech.2022-10165",
issn="1990-9772",
url="https://www.isca-speech.org/archive/pdfs/interspeech_2022/stafylakis22_interspeech.pdf"
}