Correlates of Psychological Dimensions in Talker Similarity

Abstract

This investigation identifies correlates of psychological dimensions in talker similarity. Twenty adult male talkers recorded a monosyllabic word, and 13 acoustic measurements were made from spectrograms of each talker’s production. All possible pairs of voices were presented to 11 adult listeners for similarity judgments via a paired-comparison paradigm. A four-dimensional INDSCAL analysis of the similarity ratings was employed to derive psychological dimensions of talker similarity. Correlations between the acoustic measurements and the INDSCAL dimensions revealed that fundamental frequency and word duration were moderately correlated with two of the psychological dimensions. The other two dimensions were not convincingly correlated with any of the acoustic measurements, but are best described as representing voice quality and talker age. A listener’s familiarity with the talkers did not seem to influence his judgment of voice similarity.

Perceptual similarity of regional dialects of American English

“…These findings are consistent with previous research on the perception of indexical properties of speech which has consistently found that both experienced and naive listeners can make explicit judgments about properties of unfamiliar talker's voices, including voice quality (Kreiman et al, 1994;Remez et al, 2004;Walden et al, 1978), gender (Lass et al, 1976), and sexual orientation (Avery and Liss, 1996;Gaudio, 1994), and suggest that social group properties such as regional, ethnic, and social dialect are also integral components of speech perception and spoken language processing. Dialect variation is a significant source of variability in speech and naive listeners can make explicit judgments about this source of variation in the speech of unfamiliar talkers.…”

“…In one early study of the perception of talker similarity, Walden et al (1978) asked naive listeners to rate the similarity of pairs of male talkers based on single word utterances using a seven-point similarity scale. A multidimensional scaling analysis of the results revealed two primary dimensions of within-gender talker similarity, word duration and F0.…”

Learning to recognize talkers from natural, sinewave, and reversed speech samples.

“…Differences in fundamental frequency across talkers, or variation in fundamental frequency within an utterance, do not substantially alter the consonants and vowels (although producing certain phonemes can affect fundamental frequency; see Silverman, 1986). It is not surprising, then, that vocal pitch is an extremely salient component of vocal quality and accounts for most of the variance in multidimensional scaling studies of talker recognition (Carterette & Barnebey, 1975;Gelfer, 1988;Matsumoto, Hiki, Sone, & Nimura, 1973;Voiers, 1964;Walden, Montgomery, Gibeily, & Prosek, 1978). Other impressions of vocal quality, such as the degree of breathiness, hoarseness, or creakiness, arise from spectral effects of different modes of laryngeal vibration or from morphological variations in the vocal folds.…”

The perceptual structure of pathologic voice quality

“…Further, the small number of voices studied has limited the number of dimensions that may be extracted from a set of data, so that half or more of the variance in the underlying ratings may remain unexplained ͑Murry et Nieboer et al, 1988͒. Other sampling restrictions further limit the generalizability of previous studies. Many published reports have used normal voices ͑e.g., Matsumoto et al, 1973;Walden et al, 1978;Murry and Singh, 1980;Gelfer, 1993͒, but pathologic voices have been studied very little. Many studies that did examine pathologic voices included only a single disordered population ͓e.g., tracheoesophageal speakers ͑Nieboer et al., 1988͒, hoarse voices ͑Isshiki andTakeuchi, 1970; an exception is Hammarberg et al, 1980͔ or focused on specific aspects of voice quality ͑Kreiman et al, 1994͒.…”