Syllable Lattice Based Re-Scoring For Speaker Verification

Author

Jin, Minho ; Soong, Frank K. ; Yoo, Chang D.

Author_Institution

Dept. of EECS, Korea Adv. Inst. of Sci. & Technol.

Volume

1

fYear

2006

fDate

14-19 May 2006

Abstract

The Gaussian mixture based GMM-UBM approaches have shown good performance in speaker verification without using contextual information. In this paper, we exploit the information provided in the arcs of a decoded syllable lattice for speaker verification. The forward algorithm is used to summarize this information in the syllable lattice instead of the best decoded string. The performance is evaluated on a Mandarin Chinese database. With two minutes of target speaker´s enrollment data, the proposed algorithm shows 1.03% of equal-error rate for short input utterances with an average duration of two seconds. By combining with the GMM-UBM, the system shows a 0.74% of equal-error rate

Keywords

Gaussian processes; speaker recognition; Gaussian mixture; Mandarin Chinese database; contextual information; decoded syllable lattice; forward algorithm; speaker verification; syllable lattice; Asia; Error analysis; Hidden Markov models; Lattices; Maximum likelihood decoding; Maximum likelihood linear regression; Natural languages; Speech; Statistical analysis; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech and Signal Processing, 2006. ICASSP 2006 Proceedings. 2006 IEEE International Conference on

Conference_Location

Toulouse

ISSN

1520-6149

Print_ISBN

1-4244-0469-X

Type

conf

DOI

10.1109/ICASSP.2006.1660172

Filename

1660172