A Flexible Scheme for Assigning Timing and Pitch To Synthetic Speech

A flexible scheme for assigning timing and pitch to synthetic speech is described. The scheme is designed to provide sufficiently adequate and natural control of the prosody of utterances for people to be able to use the speech which it generates as a simple but effective tool for man-machine interaction. The basic prosodic features of the utterance must be marked in the phonemic input string, while more global features are defined by parameters which can be changed as and when desired. This combines a simple input format with the capability of synthesizing many of the nuances encountered when connected passages are spoken in a rich and natural way. Since the scheme has been implemented on a computer, many decisions about precise details of pitch and timing have had to be made, usually in the absence of proper experimental justification. Each point of detail is discussed in the paper, so that the procedure is described sufficiently fully to enable it to be replicated in other research laboratories.