Towards More Intelligent Speech Interface

The demand on information increases continuously. The use of computers has been changed from application and program development more and more to the communications and data retrieval. Multimedia interface of computers allows on-line collaboration not only in an office but even in extreme situations like driving, sports, entertainment activities etc. One of the principal modality in multimedia communications is speech. Recent research in speech recognition are devotes to continuous speech recognition. Basic algorithms for “word spotting” are slightly substituted by more sophisticated “topic spotting” methods. In the opposite direction of communications from computer to the user, the speech synthesis seems to be an easier task than speech recognition. However, naturalness of the artificial speech is a complex problem. An enormous effort has been made on speech prosody. Sinusoidal and noise models of speech seem to be a promising topic for artificial speech manipulation. Keywords-mulimodal interface, speech synthesis, speech quality.