ASA 127th Meeting M.I.T. 1994 June 6-10

2pSP48. A database of environmental sounds.

Suzanne E. Boyce

Dept. of Elec., Comput., & Systems Eng., Boston Univ., Boston, MA

Richard Goldhor

AudioFile, Inc., Lexington, MA 02173

A database of 272 common environmental sounds, recorded for use in the development of a recognition system is described. These recordings represent 5--8 repetitions of 50 individual types of sound. The repetitions span a variety of recording conditions such as microphone positions, location of the source in the recording space, etc. The types of sound include airplanes, sirens, whistles, pouring small objects, telephones, motors, household appliances, footsteps, animal vocalizations, and speech. Simultaneous near- and far-field recordings were made of each sound. Unlike many commercially available collections, our database was produced via a consistent format of equipment use, equipment type, standardized microphone positions, etc. all of which has been exhaustively documented. Beyond those features mentioned above, the database includes information on the proportions and salient features of the recording spaces, the dimensions and salient features of the sources, and details of the recording and computer entry process. Currently, the database is comprised of four parts: (1) digitized audio waveforms; (2) information and notes on recording setup, recording space, source type, equipment, etc.; (3) sketches indicating shape and proportions of the recording setup, and (4) photographs showing sound sources and recording setups. This information is available in machine-readable format on a personal computer. [Work supported by NIH.]