A New Speaker Change Detection Method For Two-Speaker Segmentation

DMCA

A New Speaker Change Detection Method For Two-Speaker Segmentation (0)

Cached

Download Links

by André G. Adami , Sachin S. Kajarekar , Hynek Hermansky

Venue:	Proc. of IEEE ICASSP, Volume 4, IV-3908
Citations:	18 - 0 self

BibTeX

@INPROCEEDINGS{Adami_anew,
    author = {André G. Adami and Sachin S. Kajarekar and Hynek Hermansky},
    title = {A New Speaker Change Detection Method For Two-Speaker Segmentation},
    booktitle = {Proc. of IEEE ICASSP, Volume 4, IV-3908},
    year = {},
    pages = {3908--3911}
}

OpenURL

Abstract

In absence of prior information about speakers, an important step in speaker segmentation is to obtain initial estimates for training speaker models. In this paper, we present a new method for obtaining these estimates. The method assumes that a conversation must be initiated by one of the speakers. Thus one speaker model is estimated from the small segment at the beginning of the conversation and the segment that has the largest distance from the initial segment is used to train second speaker model. We describe a system based on this method and evaluate it on two different tasks: a controlled task with variations in the duration of the initial speaker segment and amount of overlapped speech and 2001 NIST Speaker Recognition Evaluation task that contains natural conversations. This system shows significant improvements over the conventional system in absence of overlapped speech on the controlled task.

Keyphrases

two-speaker segmentation new speaker change detection method overlapped speech speaker model second speaker model significant improvement conventional system new method natural conversation nist speaker recognition evaluation task controlled task speaker segmentation initial segment initial estimate important step prior information different task small segment initial speaker segment

DMCA