Papers
arxiv:2611.18529

eDIAR: Training-Free Open-Speaker Streaming Diarization for Long Meetings and Large Archives via e-means

Published on Nov 13
· Submitted by
Kyudan Jung
on Nov 16
Authors:
,
,
,
,

Abstract

A streaming speaker-diarization method removes the fixed speaker-slot limit by clustering speaker embeddings within slots after all slots are used, allowing unlimited speakers and efficient large-scale linking.

Meetings last up to an hour, and nobody knows in advance how many people will speak. Neural diarizers with a fixed number of output slots, such as Streaming Sortformer and Nemotron-3, are accurate and run online, but they cannot name more speakers than they have slots. We remove this limit without retraining. The slot head keeps deciding who is active and when speech overlaps, and speaker embeddings are clustered inside a single slot, with the official e-means algorithm, only after every slot has been used. In the streaming system, re-clustering is confined to a bounded window, labels are revised for a few seconds, and identities are anchored to the slots of the head. The resulting system has no limit on the number of speakers and stays accurate when speakers outnumber the slots, where fixed-slot models break down. On meetings that grow to an hour it stays far more accurate than Streaming Sortformer, which accumulates errors, and its cost per update does not grow with the meeting. The same recipe turns the four-slot Streaming Sortformer into an open-speaker system. Across recordings, e-means makes it affordable to link the speakers of a large archive frame by frame, a task that arises when conversational corpora are prepared, and it reaches the clustering quality of Lloyd's k-means several times faster. The method needs little tuning, and e-means runs with its default recipe from a few hundred to more than a million embeddings.

Community

Paper submitter

We take the speaker embeddings from each slot (WeSpeaker ResNet34) and split them using Meta's recently released e-means.
For streaming, each chunk reclusters only the most recent 90-second window and allows label corrections only within the last 10 seconds. This keeps the computational cost constant regardless of meeting length, at about 57 ms per update.
The results:
With 20 speakers, DER drops from 66.5% to 43.7%, a relative reduction of about 34%.
At the 45-minute mark of one-hour meetings with 12–20 speakers, Streaming Sortformer reaches around 57% DER, while our method stays around 25%.
When e-means is used for frame-level speaker association, matching the same speaker across recordings with 1,251 speakers is several times faster than with Lloyd's k-means.

Sign up or log in to comment

Models citing this paper 1

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.18529 in a dataset README.md to link it from this page.

Spaces citing this paper 1

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.