The Design And Evaluation On The Platform Of Speech User Interface

Speech is one of the most natural and pervasive communication modality in our everyday lives. So once the computer came to being into, the conversation between people and computer has been became a dreamful goal in the computer domain. Speech User Interface is the specific science to solve this problem. It is an important component to multi-modal Human Computer Interaction. Traditional research in speech-based technology is concentrated on speech recognition, speech synthesis and natural language understanding. But it is hardly to realize the real speech interaction rely only on these speech technologies. So we also must make researches on the communicative mechanism, Interface Model, Interaction Primitive and the design guideline of the Speech Modality.In our research, we held a Speech Model firstly; then we design and develop a test bed of Speech Interface according to the model, this is a prototype named homedoctor that people can get medical information through mouse, command and control language or by conversation with computer, through that test bed we validate and improve my model, attempt the developing methods of speech applications and learn the characteristics of speech interface .Through the test bed design and test, we validate the feasibility of Speech Interface. Based on the experiments and the problems encountered in the design and development, we summarize some design guidelines and experiences. But the Speech Interface can not broadly employ now because of the limitation of speech technologies, they are imperfect. Whereas the Speech Interface can apply in some specific domains and situations, such as the weather forecast on the telephone. In the future, the major stream of Speech Interface study and development will focus on the Multimodal Integration.