Modelling man-computer oral dialogue in noisy environment

A model of a voice controlled system will be presented in the paper. The behaviour of the system is modelled by a finite state process. The problem domain is supposed to be well bounded and very limited. An isolated word classification method is used for the recognition of user's control command or sequence of commands. A speech synthesizer is used to implement acoustic feedback control. As an illustration example its implementation and application for searching and updating database is described. Problems involved in classification and communication between the system and the user in noisy environment are treated. Optimization tradeoffs are proposed.