ISCA Archive ICSLP 2000
ISCA Archive ICSLP 2000

Hands-free human-machine dialogue - corpora, technology and evaluation

Frank K. Soong, Eric A. Woudenberg

In this paper we will review the progress of hands-free, Voice User Interface (VUI) research work at Bell Labs, including: a multichannel data base collection, technology development, and performance evaluation. Thirty-channel, simultaneous recordings have been conducted in a moving car, collecting speech from 57 subjects under various weather, road, and noise conditions. These are being used for both testing and adaptation purposes. Technology issues relevant to hands-free VUI are specifically addressed, including: (1) acoustic echo cancellation (AEC) and near-end (user's) speech detection; (2) background noise estimation and suppression; (3) reliable and timely barge-in; (4) signal pickup improvement using intelligent microphone arrangements; and (5) speaker and environment adaptation. An evaluation of the developed technologies using the car database is presented. An all software, hands-free, full duplex voice user interface demo has been implemented on a LINUX PC. The real-time demo provides services like: voice-dialing (dialing a person by name or a connected digit string), information service (accessing headline news, weather reports, sports and stock quotations), personal message service (retrieving email, voice mail and fax) and voice control of a DVD-player (selecting topics, controlling volume and video playback speeds).


Cite as: Soong, F.K., Woudenberg, E.A. (2000) Hands-free human-machine dialogue - corpora, technology and evaluation. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 4, 41-44

@inproceedings{soong00_icslp,
  author={Frank K. Soong and Eric A. Woudenberg},
  title={{Hands-free human-machine dialogue - corpora, technology and evaluation}},
  year=2000,
  booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
  pages={vol. 4, 41-44}
}