Editor:	Real World Computing Partnership Real Acoustic Environments Working Group
Copyright(c) 1998-2001	Mitsubisi Research Institute,Inc. Satoshi Nakamura, and Takanobu Nishiura, ATR Spoken Language Translation Research Laboratories. Futoshi Asano, Electrotechnical Laboratory. Takeshi Yamada, University of Tsukuba Tetsunori Kobayashi, Waseda University Nippon Telegraph and Telephone Corporation
Produced by	Mitsubishi Research Institute,Inc.

1. Objective

In recent years, processing sound signal in a real acoustic environment has become a subject of research. The importance of such research is increasing in varied fields of application such as man-machine interface, speech recognition, environmental sound recognition, echo canceller, active noise control and the like. However, no standard database is presently available.

A sound scene database in real acoustic environment will contribute to core research as a standard for objective comparison and evaluation of research results. Though a microphone array can measure factors in a three-dimensional acoustic environment, using equipment with multiple signal channels is costly and difficult. For that reason a common database is valuable in that it would allow research to be conducted even without the suitable facilities.

The RWCP Sound Scene Database in Real Acoustic Environment is a common database for research use that has been developed by the Real Acoustic Environment Subgroup of RWCP Intellectual Resource for Research and Development Working Group in Real World Computing Partnership, as a work of Real World Computing Program. The data were recorded from 1998 to 2000.

2. Contents

The content includes speech measurement data collected by microphone array and dry source of non-speech sound. These data are essential for sound scene research using a microphone array in a real acoustic environment. The content also includes explanation of algorithms related to measurements, software using the algorithms and a study case using these measurement data.

Data stored

• Non-speech sound dry sources (Hiyane,Iio@MRI)
  • non-speech sounds in unechoic room (dry sources)
  • reconstructed signals in various rooms (convolution with impulse response)
• Impulse response and speech data with microphone array (Nakamura, Yamada, Nishiura@ATR) [Homepage]
  • impulse response of fixed sound sources
  • speech data of fixed sound sources
  • impulse response of moving sound sources
  • speech data of moving fixed sound sources
  • diffusible sound source, background noise
• Impulse response at near field with microphone array (Asano@ETL) [Homepage]
• HTRF of interactive robot (Kobayashi@Waseda Univ.) [Homepage]

Algorithms and Softwares

• Measurement of impulse response using TSP(Time-Stretched Pulse) (Asano@ETL) [Homepage]
• Measurement of reverberation time (Nakamura, Nishiura@ATR) [Homepage]
• Sound source localization and separation at near fields (Asano@ETL) [Homepage]
• Sweeping Echo (an echo whose frequency rises with time) (Kiyohara,Furuya,Kaneda@NTT)
• Convolution between dry source and impulse response (Hiyane,Iio@MRI)
• Display and manipulation of sound signals

Application research cases

• Non-speech sound recognition with microphone array (Hiyane,Iio@MRI) [Homepage]
• Sound source discrimination and speech recognition based on HMM using a microphone array (Nakamura, Nishiura@ATR) [Homepage]
• Voice Activity Detection in Noisy Environments (Yamada@Tsukuba Univ.)

3. Distribution and How to get

The database may be put to unlimited use in research work and other applications such as product development. It is primarily intended for research and technological development of non-speech sound and microphone array.

Usage is subject to the following conditions.

• All rights of data, documents and software contained in the database are reserved by each copyright holder.
• Integration in a product of data contained in the database must be accompanied by a written permission from the copyright holder.
• The database must be credited when a result obtained by using the database is published in a document, web site, newspaper and the like, as follows:
  "The present study was conducted using the XXX data of the RWCP Sound Scene Database in Real Acoustic Environment."
• Real World Computing Partnership and copyright holders shall have no liability for any damage of any kind suffered by a user or a third party, whether direct or indirect, by using the present database.
• The database may not be used, either in its entirety or part thereof, for redistribution, reproduction, conversion, broadcast, screening and rental without the written permission of the copyright holder.
• The software may be subject to other distribution conditions. Refer to the page of each software program.

The database is contained in three DVD-ROMs. To obtain a copy, please refer to the section entitled "How to obtain DVD-ROM" in this guide. Distribution is limited in number.

4. Inquiries

For details regarding application and content of the DVD-ROM, please make your inquiries after reading this page thoroughly so as to help facilitate our response to your inquiry. If you have questions and comments, please email to sound-db@mri.co.jp.

5. Links

• RWCP - Real World Computing Partnership
• RWC Databases

6. Publication

• S.Nakamura, K.Hiyane, F.Asano, T.Nishiura, and T.Yamada: "Acoustical Sound Database in Real Environments for Sound Scene Understanding and Hands-Free Speech Recognition", 2nd International Conference on Language Resources & Evaluation, Athen (2000.6).
• S.Nakamura, K.Hiyane, F.Asano, and T.Endo: "Sound scene data collection in real acoustical environments", J. Acoust. Soc. Japan (E) Vol.20, No.3 (1999.5).
• K.Hiyane, S.Nakamura, F.Asano, and T.Endo: "Real World Speech and Acoustic Database", Proc. of IEICE, 1999.3 (in Japanese).
• S.Nakamura, K.Hiyane, F.Asano, and T.Endo: "Real World Sound Scene Database", Proc. of Acoustical Society of Japan, 1998,9 (in Japanese).
• S.Nakamura, K.Hiyane, F.Asano, and T.Endo: "Sound Scene Database in Real Acoustic Environments", Oriental COCOSDA Workshop '98(1998.5).