Linker improves speaker tracking beyond fixed diarizer limits
Slots Are Not Speakers: Training-Free Speaker Linking for Frozen Streaming Diarizers
SoundArtificial Intelligence
Summary
Many voice diarizers can only track a fixed number of speakers at once, which causes confusion when more people talk. The authors created a new way to connect a diarizer’s speaker slots with actual speakers without needing extra training. Their method splits and matches speech segments to known speakers, cutting confusion without affecting detection accuracy. It works well even in challenging scenarios like courtroom hearings with many speakers.
What this means in practice
- •For conference systems developers: Improve identification of individual speakers in live meetings when more people than diarizer slots speak.
- •For legal transcription teams: Enhance accuracy in tracking many speakers simultaneously during courtroom hearings without retraining diarization models.
Authors
Aditya Srinivas Menon, Kumarmanas Nethil, Kavya Manohar
Abstract
This paper presents a causal, training-free linker that allows a frozen streaming diarizer to track more speakers than it has output slots. Streaming end-to-end diarizers such as Streaming Sortformer emit a fixed number N of concurrent speaker slots. Above N speakers, they keep detecting speech but reuse slots across people, because N bounds how many speakers the model can express at once, not how many it will meet. The linker uses only the host's slot posteriors and the audio. It keeps the frames where one slot is active, splits each slot where the speaker changes, embeds each piece with an off-the-shelf speaker encoder, and matches it against a bank of speakers seen so far. It only relabels, so missed speech and false alarm are identical to the host's and every gain is a reduction in speaker confusion. On 1418 sessions from eight corpora, it reduces the 4-slot host's DER above its cap from 36.8% to 22.6%, and barely moves it below the cap. The same constants transfer to the host's 8-slot successor and to two streaming EEND hosts from a different model family. On courtroom hearings with up to 23 speakers it reduces cpWER from 82.7% to 24.4%.