Twenty listeners rated 10 speakers from the Air Force Cambridge Research Laboratory’s (AFCRL) speaker library, using a 12-item rating form developed from Osgood’s (1957) semantic differential rating form techniques. Physical characteristics of the same speakers' voices also were measured. The results revealed that speakers can be reliably differentiated on the basis of (1) listener judgments alone, (2) listener judgments combined with physical voice measures, and (3) physical voice measures alone. Speakers' voice characteristics, as determined by listeners' judgments alone, can be accounted for by two basic factors which account for over 90% of the variance. Factor analysis of both judged and actual physical voice measures led to a 5-factor solution accounting for more than 90% of the total variance.

No abstract

This paper represents an effort to update and expand an earlier paper, "Speech Analysis, Synthesis, and Processing-A Selected Bibliography,' prepared and published by this author at Texas Instruments Inc., Dallas, in 1963. Selections included in this paper will provide the research with a fairly extensive and representative presentation of relevant source material in the areas to which the title refers.

No abstract

A study was performed to investigate the definition of a perceptual space within which listeners locate voices, to the end that the effects o[ manipulating speaker, hardware, and listener characteristics can be measured, and, eventually, that specifications for elements of a voice-communications system might be prepared to produce the desired system characteristic. In the experiments reported, taped speech samples were rated by listeners, using Osgood's semantic differential method. Previous study indicated only four basic dimensions were required to account for ratings given speakers on a large number of characteristics. In a second experiment, a reduced number of characteristics, selected from the original list as best representing the four necessary factors, was used by listeners to rate recorded speaker samples. The experimental design allowed examination of the effects on ratings due to differences between listeners, due to repetition of the rating task, and to order of speaker presentation. Results of these examinations and the following are presented: (1) adequacy of the original factors in accounting for listeners' ratings; (2) differentiation between speakers; (3) reliability of ratings; and (4) familiarity of previously unheard voices. [This research was supported by AFCRL (OAR) under Contract AF19(628)-345.]