15th Annual Conference of the International Speech Communication Association

September 14-18, 2014

Listener Estimation of Speaker Age Based on Whispered Speech

Angelika Braun, Daniela Decker

Universität Trier, Germany

The assessment of a given speaker's age has been shown to rest on laryngeal as well as vocal tract features. This study attempts to separate these two elements by comparing listener performance based on whispered and phonated speech. A total of 45 speakers belonging to three different age groups (20–35; 45–60; 70 and above) were assessed by 20 listeners. Results show that although listener performance decreases with whispered speech, it is still well above chance level. No significant difference was found between age groups, but female speakers were assessed correctly more often than males in both conditions. There are forensic implications to this finding, since in that context, age estimation sometimes has to be carried out on whispered speech, e.g. if whispering is used as a voice disguise.

Full Paper

Bibliographic reference.  Braun, Angelika / Decker, Daniela (2014): "Listener estimation of speaker age based on whispered speech", In INTERSPEECH-2014, 1712-1716.