2D Audiovisual Text-to-Speech Synthesis for Human-Machine Interaction in Dutch

Speech has always been the most important means of communication between humans. Therefore, using speech in machine-human communication can help in increasing the naturalness of the communication between a computer system and a user. Systems that can make a machine pronounce any given input text are referred to as text-to-speech systems. To further enhance the communication, a talking head can be added to the text-to-speech synthesis, since the addition of this synthetic visual speech mode will improve the intelligibility of the artificial speech (Pandzic et al., 1999). Furthermore, users will perceive this multimodal speech communication as more natural and they will feel more positive and confident if they can see the (artificial) person that is talking to them. In this paper we propose an audiovisual text-to-speech synthesis system for Dutch that is able to create both the target auditory and the target visual speech by using a same audiovisual database, which makes it possible to maximize the intermodal coherence in the audiovisual output signal.