Engineering Paralinguistics: And Next … the Transparent Speaker?

Dr.-Ing. cand. habil. Björn Schuller
(Technische Universität München)

Recently, an increasing number of speaker states and traits is adressed in
research on automatic speaker classification. Examples comprise
personality traits, likability, height, and intoxication of a person derived
from characteristics of the voice and the spoken content. This talk aims to
provide an overview on the dominant methodology used, benchmark
accuracies reached as manifested by research Challenges the speaker
held, and concludes with recent trends in the field and new avenues to
overcome data sparseness and unreliability.

Der Vortrag im Zentrum für interdisziplinäre Linguistik (ZefiS) wurde am 21. Dezember 2011 aufgezeichnet. Ansprechpartnerin im ZefiS ist Prof. Dr. Monika Rathert.