Subjective age estimation using speech sounds: Comparison with facial images

Abstract

We have defined the perception of one's own age as "subjective age", and have so far approached using other people's facial images. In this paper, we propose a relative estimation method for subjective age by using people's speech sounds and their chronological age. A relative estimation task is performed, wherein an estimator gives rating values to other people about whether they are older or younger than the estimator. In this task, the difference in the actual age between the estimator and the rating value of the person are measured. We plot the results on a two-dimensional plane with the x-axis as the relative age and the y-axis is the estimation result. Thus, the distribution with the plotted points of upper-right direction is obtained, which is approximated by a logistic function. The zero crossing point in the approximation curve with the x-axis is defined as the shift value in the subjective age. 57 total estimators, including 28 males and 29 females from 25 to 44 years old participated in this experiment. As a result, the subjective age using speech sounds tended to be older than using facial images. The tendency was also more remarkably visible in the male groups than in the female group. However, variance was relatively higher in the male young-middle (35-44) and the female young (25-34) groups than using facial images, which indicated that the subjective age varies according to the profile of estimators, such as age and gender.

Age-Related Bias in Age Estimation Based on Facial Images of Others

“…Concerning other channel types, we reported in our previous study that people also exhibit age estimation biases for voices, but the opposite bias occurs. People tend to overestimate the age of their own voice compared to others' voices (Nishimoto, Azuma, Miyamoto, Fujisawa, & Nagata, 2008). On the other hand, in the context of computer-mediated interaction, text or facial images are very often the only types of information we see about other people and them about us, in social networking service (SNS) communities.…”

Predictive biometrics: a review and analysis of predicting personal characteristics from biometric data

“…The investigation of age prediction from gait data can be found in [44,105], while [46,106] investigate the prediction of age from iris data using different sets of geometric and texture features. In [45,100,107], the authors have investigated the prediction of age from voice data, and Merkel et al [108] examines the prediction of subject age from the fingerprint. Predictive accuracies range from a minimum of 57% (for the iris modality) to a maximum of 77% (for the signature modality), but the problems noted above make drawing specific conclusions unwise.…”

“…For example, the notion of exploiting 'soft biometrics' (biometric characteristics which are specific to an individual; however, not in themselves unique -subject age, e.g. or gender) is not new, but has gained in prominence, both as a means of supplementing unique biometric data to improve identification processes and as a way of determining additional information about individuals or particular application scenarios which may prove useful in specific contexts [37][38][39][40][41][42][43][44][45][46][47].…”

Age Prediction from Iris Biometrics

“…Biometric-based age estimation is perhaps the most common and valuable manifestation of this process. The literature shows that face [1,2], speech [3,4], and signature [5] biometrics, have received most attention in the research area of age prediction. However, even though it is thus possible to find some interesting work dealing with age estimation based on various different biometric modalities; there is only one reported study [6] concerning age estimation based on iris biometrics (although gender and ethnic group prediction from iris images is proposed in [7,8] and [8,9] respectively).…”