Towards spontaneous speech recognition for on-board car navigation and information systems

Speech recognition is seen to be of great benefit in on-board car navigation systems and assistance. The command word approach will be used for applications in the near future since the small active vocabulary and the hierarchical structure is much easier to cope with, from the developers’ side. An alternative approach, using spontaneous speech input, is far more complex but provides the user with an interface that is very intuitive and has fewer restrictions. The user can rely upon his or her experience in inter-human communication and utter spontaneous queries. In this paper, we describe the requirements and the collection of a continuous car speech data base and show first recognition results obtained under different environmental conditions in the car.