ISIS and NISIS: New bilingual dual-channel speech corpora for robust speaker recognition

Amita Pal; Smarajit Bose; Mandar Mitra; Sandipan Roy

ISIS and NISIS: New bilingual dual-channel speech corpora for robust speaker recognition

Amita Pal, Smarajit Bose, Mandar Mitra, Sandipan Roy

Department of Mathematical Sciences

Research output: Chapter or section in a book/report/conference proceeding › Chapter in a published conference proceeding

1 Citation (SciVal)

Abstract

It is standard practice to use benchmark datasets for comparing meaningfully the performance of a number of competing speaker identification systems. Generally, such datasets consist of speech recordings from different speakers made at a single point of time, typically in the same language. That is, the training and test sets both consist of speech recorded at the same point of time in the same language over the same recording channel. This is generally not the case in real-life applications. In this paper, we introduce a new database consisting of speech recordings of 105 speakers, made over four sessions, in two languages and simultaneously over two channels. This database provides scope for experimentation regarding loss in efficiency due to possible mismatch in language, channel and recording session. Results of experiments with MFCC-based GMM speaker models are presented to highlight the need of such benchmark datasets for identifying robust speaker identification systems.

Original language	English
Title of host publication	Proceedings of the 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012
Subtitle of host publication	Volume 2
Pages	936-939
Number of pages	4
Publication status	Published - 1 Dec 2012
Event	2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012 - Las Vegas, NV, USA United States Duration: 16 Jul 2012 → 19 Jul 2012

Conference

Conference	2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012
Country/Territory	USA United States
City	Las Vegas, NV
Period	16/07/12 → 19/07/12

Keywords

Classification accuracy
Gaussian mixture models
Mel frequency cepstral coefficients
Robust speaker recognition

ASJC Scopus subject areas

Computer Graphics and Computer-Aided Design
Computer Vision and Pattern Recognition

Cite this

ISIS and NISIS: New bilingual dual-channel speech corpora for robust speaker recognition. / Pal, Amita; Bose, Smarajit; Mitra, Mandar et al.
Proceedings of the 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012: Volume 2. 2012. p. 936-939.

Research output: Chapter or section in a book/report/conference proceeding › Chapter in a published conference proceeding

Pal, A, Bose, S, Mitra, M & Roy, S 2012, ISIS and NISIS: New bilingual dual-channel speech corpora for robust speaker recognition. in Proceedings of the 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012: Volume 2. pp. 936-939, 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012, Las Vegas, NV, USA United States, 16/07/12.

@inproceedings{24f2958f5f274d8da949f1bcac64d2f2,

title = "ISIS and NISIS: New bilingual dual-channel speech corpora for robust speaker recognition",

abstract = "It is standard practice to use benchmark datasets for comparing meaningfully the performance of a number of competing speaker identification systems. Generally, such datasets consist of speech recordings from different speakers made at a single point of time, typically in the same language. That is, the training and test sets both consist of speech recorded at the same point of time in the same language over the same recording channel. This is generally not the case in real-life applications. In this paper, we introduce a new database consisting of speech recordings of 105 speakers, made over four sessions, in two languages and simultaneously over two channels. This database provides scope for experimentation regarding loss in efficiency due to possible mismatch in language, channel and recording session. Results of experiments with MFCC-based GMM speaker models are presented to highlight the need of such benchmark datasets for identifying robust speaker identification systems.",

keywords = "Classification accuracy, Gaussian mixture models, Mel frequency cepstral coefficients, Robust speaker recognition",

author = "Amita Pal and Smarajit Bose and Mandar Mitra and Sandipan Roy",

year = "2012",

month = dec,

day = "1",

language = "English",

isbn = "9781601322258",

pages = "936--939",

booktitle = "Proceedings of the 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012",

note = "2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012 ; Conference date: 16-07-2012 Through 19-07-2012",

}

TY - GEN

T1 - ISIS and NISIS

T2 - 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012

AU - Pal, Amita

AU - Bose, Smarajit

AU - Mitra, Mandar

AU - Roy, Sandipan

PY - 2012/12/1

Y1 - 2012/12/1

N2 - It is standard practice to use benchmark datasets for comparing meaningfully the performance of a number of competing speaker identification systems. Generally, such datasets consist of speech recordings from different speakers made at a single point of time, typically in the same language. That is, the training and test sets both consist of speech recorded at the same point of time in the same language over the same recording channel. This is generally not the case in real-life applications. In this paper, we introduce a new database consisting of speech recordings of 105 speakers, made over four sessions, in two languages and simultaneously over two channels. This database provides scope for experimentation regarding loss in efficiency due to possible mismatch in language, channel and recording session. Results of experiments with MFCC-based GMM speaker models are presented to highlight the need of such benchmark datasets for identifying robust speaker identification systems.

AB - It is standard practice to use benchmark datasets for comparing meaningfully the performance of a number of competing speaker identification systems. Generally, such datasets consist of speech recordings from different speakers made at a single point of time, typically in the same language. That is, the training and test sets both consist of speech recorded at the same point of time in the same language over the same recording channel. This is generally not the case in real-life applications. In this paper, we introduce a new database consisting of speech recordings of 105 speakers, made over four sessions, in two languages and simultaneously over two channels. This database provides scope for experimentation regarding loss in efficiency due to possible mismatch in language, channel and recording session. Results of experiments with MFCC-based GMM speaker models are presented to highlight the need of such benchmark datasets for identifying robust speaker identification systems.

KW - Classification accuracy

KW - Gaussian mixture models

KW - Mel frequency cepstral coefficients

KW - Robust speaker recognition

UR - http://www.scopus.com/inward/record.url?scp=84873303777&partnerID=8YFLogxK

M3 - Chapter in a published conference proceeding

AN - SCOPUS:84873303777

SN - 9781601322258

SP - 936

EP - 939

BT - Proceedings of the 2012 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2012

Y2 - 16 July 2012 through 19 July 2012

ER -

ISIS and NISIS: New bilingual dual-channel speech corpora for robust speaker recognition

Abstract

Conference

Keywords

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this