Speaker identification and classification of speaker characteristics The classification of speaker characteristics by a person's voice can be regarded as a classification task.
The goal is to decide whether a speakers belongs to a specific speaker group. Speaker groups in this context are persons having
specific characteristics in common like gender, age, country of birth, emotional state and so on. Even the identification of a
speaker can be regarded as a classification of speaker groups, where the group contains exactly one person, i.e., the speaker himself.
Thus, the classification of speaker characteristics can be achieved by means of speaker identification. The expected output of this
project is the development of a generic system for the classification of speaker characteristics. The system will be evaluated on the
two speaker characteristics age and language id / language of birth. The project will be concluded by a combination of the developed
system with an existing speaker identification system. | Project manager: Prof. Dr.-Ing. Elmar Nöth, Prof. Dr. Elizabeth Shriberg
Project participants: Dr. rer. biol. hum. Tobias Bocklet
Keywords: automatic speaker identification; age recognition; speaker characteristics; speech procession
Duration: 1.1.2009 - 21.12.2010
Sponsored by: BaCaTec
Mitwirkende Institutionen: Stanford Research Institute (SRI)
Contact: Bocklet, Tobias E-Mail: tobias.bocklet@cs.fau.de
|