DiaPer: End-to-End Neural Diarization With Perceiver-Based Attractors

journal article in Web of Science

English

Until recently, the field of speaker diarization was dominated by cascaded systems. Due to their limitations, mainly re- garding overlapped speech and cumbersome pipelines, end-to-end models have gained great popularity lately. One of the most success- ful models is end-to-end neural diarization with encoder-decoder based attractors (EEND-EDA). In this work, we replace the EDA module with a Perceiver-based one and show its advantages over EEND-EDA; namely obtaining better performance on the largely studied Callhome dataset, finding the quantity of speakers in a conversation more accurately, and faster inference time. Further- more, when exhaustively compared with other methods, our model, DiaPer, reaches remarkable performance with a very lightweight design. Besides, we perform comparisons with other works and a cascaded baseline across more than ten public wide-band datasets. Together with this publication, we release the code of DiaPer as well as models trained on public and free data.

Attractor, DiaPer, end-to-end neural diarization, perceiver, speaker diarization.

LANDINI, F.; DIEZ SÁNCHEZ, M.; STAFYLAKIS, T.; BURGET, L.

3. 7. 2024

1558-7916

IEEE Transactions on Audio, Speech, and Language Processing

32

7

United States of America

3450

3465

16

https://ieeexplore.ieee.org/document/10584294

landini_IEEE_Transactions_2024_DiaPer_End-to-End_Neural_Diarization.pdf