original source: http://web.archive.org/web/20060831010952/http://cslu.ece.ogi.edu/nsel/data/SpEAR_database.html

SpEAR (Speech Enhancement and Assessment Resource) Database

http://ee.ogi.edu/NSEL/ (Beta Release v1.0)

---

The SpEAR Database contains carefully selected samples of noise corrupted speech with clean speech references. This beta version is a small initial release of the database to allow feedback for future developments of this resource.

All speech and noise sources have been acoustically combined and re-recorded. Synchronous clocking is used to provide an exact time-aligned reference to the clean speech signal (see Technical Details).

Download

• zip format

Contents (more detailed description)

• Original Speech Files (TIMIT/)

• Noisy Speech Recordings (Noisy_Recordings/)- Recorded speech and recorded noise: acoustically combined and re-recorded. Various noise sources and different SNR levels.

• Lombard Speech (Lombard/) - Live speech in a noisy environment. Recorded noise sources are used to allow extraction of a cleaned speech reference.

• Monaural Recordings (Monaural/) - Two recorded speech signals: acoustically combined and re-recorded.

Future releases will contain a more substantial number of samples, additional additive noise sources, cellular and coding noise, etc.

Use of this data is free, but should be referenced in all publications as follows:

Speech Enhancement Assessment Resource (SpEAR) Database. http://ee.ogi.edu/NSEL/. Beta Release v1.0. CSLU, Oregon Graduate Institute of Science and Technology. E. Wan, A. Nelson, and Rick Peterson.

People

Principle Investigator:

• Eric Wan

Graduate Research Assistants:

• Houwu Bai

• Rudolph van der Merwe

• Alex T. Nelson (graduated 2000)