Study on Perception of Speaker Age by Semantic Differential Method

Humans can perceive ages of speakers from uttered voices by their own judgements. The perceived ages are called perceptual ages (PAs). Many earlier studies focused on statistical correlations between aging voices and acoustic features without taking into account the fact that human perception is vague rather than precise [1]. This paper focuses on the psychological factors to study human perceptions for aging voices. An experiment was carried out to evaluate the aging voices by candidates of semantic primitives, and the results of the listening test were analyzed by Semantic Differential Method and Regression Analysis to investigate impressions that human use to estimate PAs of speakers. Results show that with regards to both male and female voices, the Metal Factor (Deep - Flimsy, Full - Delicate, Rich ? Thin, Heavy - Light), which shows a linear relation with both male and female PAs, is the most important factor that helps listeners judge PAs of uttered voices. In addition, the rest of the factors show both linear and non-linear relationships with male aging voices, while only non-linear relations with female aging voices