Speech as a Source for Ubiquitous User Modeling

Christian Müller, Frank Wittig

In: Proceedings of the 9th International Conference on User Modeling (UM 03), Johnstown, USA, 2003.. International Conference on User Modeling (UM-03) June 22-26 University of Pittsburgh, Johnstown PA United States 2003.


In this paper, we present an approach on how to use speech as a source for user modeling in a mobile and ubiquitous context. In particular, we exploit different abstraction levels of speech features to estimate the user's age and gender. To solve the classification task, we compared several well known machine learning techniques such as artificial neural networks and support vector machines. The results of our study imply that one can indeed successfully extract higher level information from the raw speech data. We show how this approach is integrated into a generic resource adaptive system architecture. One particular instance of this system is an implementation of a mobile pedestrian navigation system.

muellerWittig2003.pdf (pdf, 313 KB )

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz