Series in Microelectronics
George S. Moschytz
Speaker Separation and Tracking.
2006; VIII, 144 pages. € 64,00. ISBN 3-86628-054-8
Today, social interaction analysis is conducted manually by observers, by interviews or is restricted to specially equipped rooms. During social interactions, oral communication is an important, if not the most important communication modality. For example, speech is the main communication modality during discussions, meetings and conferences. In this work, a tool is developed which automatically annotated oral communication as part of social interactions.
The system design is induced by the vision of a tool which helps to annotate oral communication anytime and anywhere. Therefore, novel design constrains for a speaker separation and tracking system are considered. These constraints imposed by the mobile application are limited number of sensors and mounting positions, limited computation power, dynamic environment and simultaneous speech. To address these constraints, a two-step approach is proposed which combines blind source separation and speaker tracking. To evaluate different configurations, a benchmark methodology for two-step SSaT systems is introduced, which considers speaker recognition and computational load.
Urs Anliker received the Dipl.-Ing. (MSc) degree in electrical engineering from ETH Zurich, Switzerland, in 2000, and the Dr. sc. ETH Zurich (PhD) degree in information technology and electrical engineering from the Swiss Federal Institute of Technology (ETH), Zurich, Switzerland, in 2005. He joined the Electronics Laboratory at ETH Zurich in 2000 as a research and teaching assistant in the Wearable Computing Group. His research interests include low power wearable system design, blind source separation and speaker identification systems.
Keywords: Blind Source Separation, Speaker Localiztation, Speaker Tracking, Mobile Application
Direkt bestellen bei / to order directly from: Hartung.Gorre@t-online.de