An investigation of speaker photograph identification

Abstract

The purpose of this investigation was to determine whether listeners were capable of speaker photograph identification based on recorded speech samples. A standard prose passage was recorded by 30 speakers, 15 females and 15 males. A master tape of all recorded readings was played to a group of 30 subjects for speaker photograph identification purposes. The subjects were shown photographic slides of two persons on a screen and simultaneously heard a tape recording of a single speaker reading the standard prose passage. Their task was to determine to which of the two photographs the speaker’s voice belonged. All subjects participated in two experimental sessions. In one session they were shown full-view photographic slides and in another, portrait-view slides. The order of presentation of the two kinds of slides was randomized so that 15 subjects were shown full-view slides first and 15 subjects were shown portrait-view slides first. Subjects were also asked to indicate the confidence of their speaker photograph identifications on a seven-point rating scale, and to describe the criteria upon which they based their decisions. Results indicate that the subjects were capable of identifying the full-view photographs of female and male speakers with slightly better than chance guessing accuracy when presented with their recorded speech samples. Implications of these findings and suggestions for future research are discussed. Subject Classification: [43]70.30, [43]70.65.