![]() |
ESCA - NATO/RSG 10 Workshop on
|
![]() |
This study describes the development of a voice-based driver interface that has been carried out in a project within the Swedish RTI program. The work covers aspects on speech recognition under adverse conditions, speech synthesis and information management. It has resulted in a pilot demonstrator used for preliminary field trials. As a test bed for a voice based operator interface, the traffic information provided by RDS/TMC (Radio Data System / Traffic Message Channel) has been used. This information is currently broadcast in the Gothenburg region of Sweden. The experimental system presents the decoded TMC messages in spoken output. Since many different messages are broadcast, it should be possible to make a selection of the most interesting and urgent ones. This may be done using an on-board, voice-controlled, PC-based information management system. A speaker adaptive recognizer for a limited vocabulary is used. The noise polluted environment in a car will impede the performance of a speech recognizer. In order to substantially reduce these effects the speech is input via a microphone array using spatial and adaptive filtering techniques. Furthermore, the recognition system is specifically designed to be able to compensate for ambient noise. The voice output is provided by a formant speech synthesizer. To improve the speech quality, phonetic text input has been used to a considerable extent combined with careful design of the spoken utterances.
Bibliographic reference. Granström, Björn / Blomberg, Mats / Elenius, Kjell / Roxstrom, A. / Nordholm, S. / Nordebo, S. / Claesson, I. / Waernulf, B. / Eke, K. (1993): "An experimental voice-based traffic information provider for vehicle drivers", In AST-1993, 79-82.