Publication Details
Ivector-Based Prosodic System For Language Identification
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT)
Ferrer Luciana (SRI)
Scheffer Nicolas (SRI)
Language Identification, Prosody, iVectors, Joint Factor Analysis
This paper is on a LID system based on prosodic features. Extraction of the pitch, energy, and duration allows us to represent the three components of prosody.
Prosody is the part of speech where rhythm, stress, and intonation are reflected. In language identification tasks, these characteristics are assumed to be language dependent, and thus the language can be identified from them. In this paper, an automatic language recognition system that extracts prosody information from speech and makes decisions about the language with a generative classifier based on iVectors is built. The system is tested on the NIST LRE09 dataset. The results are still not comparable to state-of-the-art acoustic and phonotactic systems. However, they are promising and the fusion of the new approach with an iVector-based acoustic system is found to bring further improvements over the latter.
@INPROCEEDINGS{FITPUB9997,
author = "David Gonz\'{a}lez Mart\'{i}nez and Luk\'{a}\v{s} Burget and Luciana Ferrer and Nicolas Scheffer",
title = "Ivector-Based Prosodic System For Language Identification",
pages = "4861--4864",
booktitle = "Proc. International Conference on Acoustics, Speec",
year = 2012,
location = "Kyoto, JP",
publisher = "IEEE Signal Processing Society",
ISBN = "978-1-4673-0044-5",
doi = "10.1109/ICASSP.2012.6289008",
language = "english",
url = "https://www.fit.vut.cz/research/publication/9997"
}