Segmentation Techniques in Speech Synthesis
Gordon E. Peterson, William Yang Wang, Eva Sivertsen · The Journal of the Acoustical Society of America · 1958
A basic method of speech synthesis is described in which discrete segments of recorded utterances are joined together to produce continuous speech. The segments are characterized as (a) containing parts of two phones with their mutual influence in the middle of the segment, and (b) beginning and ending at the phonetically most stable position of each phone. All segments containing the same articulatory sequence have been defined as a dyad. The method of synthesis described includes not only articulatory phones, but also intonation, stress, and duration. A large number of segments is required and various techniques of obtaining the segments for speech synthesis are discussed. The method is limited to a specific dialect, and practically it is limited to a single speaker.