Publication Details
Voice activity detection in video mediated communication from gaze
Eivazi Shahram (FIT BUT)
Bednařík Roman (University of Eastern Finland)
gaze tracking, voice activity detection, speaker recog-nition, machine learning, Support Vector Machines
This paper discuses prediction of active speaker in multi-party video mediated communication from gaze data. In the explored setting, we predict voice activity of participants in one room based on gaze recordings of a single participant in another room. The two rooms were connected by high definition and low delay audio and video links and the participants engaged in different activities ranging from casual discussion to simple casual games. We treat the task as classification problem. We evaluate different types of features and parameter setting in the context of Support Vector Machine classification framework. The results show that the speaker activity can be correctly predicted with the proposed approach in 90 % of the time for which the gaze data are available.
@INPROCEEDINGS{FITPUB9861,
author = "Michal Hradi\v{s} and Shahram Eivazi and Roman Bedna\v{r}\'{i}k",
title = "Voice activity detection in video mediated communication from gaze",
pages = "329--332",
booktitle = "ETRA '12 Proceedings of the Symposium on Eye Tracking Research and Applications",
year = 2012,
location = "Santa Barbara, US",
publisher = "Association for Computing Machinery",
ISBN = "978-1-4503-1221-9",
doi = "10.1145/2168556.2168628",
language = "english",
url = "https://www.fit.vut.cz/research/publication/9861"
}