INTERSPEECH 2008
9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Two's a Crowd: Improving Speaker Diarization by Automatically Identifying and Excluding Overlapped Speech

Kofi Boakye, Oriol Vinyals, Gerald Friedland

ICSI, USA

We present an update to our initial work [1] on overlapped speech detection for improving speaker diarization. Specifically, we describe the addition of new features and feature warping techniques that improve segmenter and, consequently, diarization performance. We also demonstrate improved diarization performance by additionally using overlap segment information in a new diarization pre-processing step which excludes overlap segments from speaker clustering. On a subset of the AMI Meeting Corpus we show that this overlap exclusion step nearly triples the relative improvement of diarization error rate as compared to overlap segment post-processing alone.

Full Paper

Bibliographic reference.  Boakye, Kofi / Vinyals, Oriol / Friedland, Gerald (2008): "Two's a crowd: improving speaker diarization by automatically identifying and excluding overlapped speech", In INTERSPEECH-2008, 32-35.