On the Ability of the Glace-Holmes Synthesizer to Produce High-Quality Speech
Weiant Wathen‐Dunn, Henry I. Soron, Sheldon B. Michaels · The Journal of the Acoustical Society of America · 1970
Formant speech synthesizers are important on two counts. The signals controlling them are related to meaningful articulatory events and are likewise very compact descriptions of speech. They are thus important for experiments on speech generation, perception, and synthesis-by-rule, and for reduced-data-rate communications. However, before such synthesizers can be used for fully meaningful investigations, it is necessary to know the inherent capability of the synthesizer to produce high-quality speech. This paper reports attempts to maximize the quality of speech produced by a Glace-Holmes synthesizer. Talkers produced utterances ranging from vowels to entire sentences. These were carefully recorded and spectrum analyzed at 10-msec intervals. The resulting data are used as a base for optimizing synthesizer output in an attempt to replicate the original speech. Speech quality judgments are made both informally and by formal listening tests. These data should prove useful in setting a standard against which the performance of communications systems or the results of speech experiments utilizing the Glace-Holmes synthesizer can be judged.