1996 Speaker Recognition Benchmark
Item Name: | 1996 Speaker Recognition Benchmark |
Author(s): | Mark Przybocki, Alvin Martin |
LDC Catalog No.: | LDC96S61 |
ISBN: | 1-58563-059-4 |
ISLRN: | 678-642-946-413-4 |
DOI: | https://doi.org/10.35111/dvct-4398 |
Member Year(s): | 1996, 1997 |
DCMI Type(s): | Sound |
Sample Type: | 1-channel ulaw |
Sample Rate: | 8000 |
Data Source(s): | telephone conversations |
Project(s): | NIST SRE |
Application(s): | speaker identification |
Language(s): | English |
Language ID(s): | eng |
License(s): |
LDC User Agreement for Non-Members |
Online Documentation: | LDC96S61 Documents |
Licensing Instructions: | Subscription & Standard Members, and Non-Members |
Citation: | Przybocki, Mark, and Alvin Martin. 1996 Speaker Recognition Benchmark LDC96S61. Web Download. Philadelphia: Linguistic Data Consortium, 1996. |
Related Works: | View |
Introduction
1996 Speaker Recognition Benchmark was developed by the Linguistic Data Consortium (LDC) and the National Institute of Standards and Technology (NIST). It contains approximately 50 hours of conversational English telephone speech collected by LDC and used in the NIST-sponsored 1996 Speaker Recognition Evaluation.
The ongoing series of SRE yearly evaluations conducted by NIST are intended to be of interest to researchers working on the general problem of text independent speaker recognition. To this end the evaluations are designed to be simple, to focus on core technology issues, to be fully supported, and to be accessible to those wishing to participate. The focus of this evaluation was on detection of the presence of a hypothesized target speaker, given a segment of conversational speech over the telephone.
This corpus represents the first yearly Speaker Recognition Evaluation conducted by NIST using approximately 4000 target trials to identify 40 target speakers.
Data
The data in this corpus is a subset of the Switchboard-1 (LDC93S7) corpus, (now superseeded by Switchboard-1 Release 2 (LDC97S62)). It consists of Development Data and Evaluation Data. Both sets include training and test segments.
The Development Data includes both training and test segments for about 45 male and 45 female speakers, comprising 1,558 files and totalling 10.4 hours. The training data consists of about four one minute segments of speech data for each target speaker. The test data contains shorter segments of speech data (three, 10, and 30 seconds) that were taken from different conversations for each speaker.
The Evaluation Data includes about 20 male and 20 female target speakers and 200 male and 200 female non-target speakers, comprising 9,394 files and totalling 39.4 . All of these speakers are different from the speakers in the Development Data set. Training data are supplied for each of the target speakers, in the same manner as the Development Data. Test data is supplied for both the target and the non-target speakers, in the same manner as the Development Data.
Here is a breakdown of the amount of files and durations contained in the corpus:
Development | Evaluation | |||
---|---|---|---|---|
Training | Test | Training | Test | |
Files | 352 | 1,206 | 160 | 9,234 |
Durations (hours) | 5.8 | 4.6 | 2.6 | 36.8 |
All files are in WAVE format with an 8 kHz sample rate.
Samples
Please listen to the following audio sample (SPH).
Updates
None at this time.