Spotting human activities and gestures in continuous data streams [Elektronische Ressource] / presented by Andreas Zinnen

Description

Spotting Human Activities and Gesturesin Continuous Data StreamsA dissertation submitted toTECHNISCHE UNIVERSITÄT DARMSTADTFachbereich Informatikfor the degree ofDoktor-Ingenieur (Dr.-Ing.)presented byANDREAS ZINNENDipl. Inform.thborn 5 of June, 1978in Bernkastel-Kues, GermanyProf. Dr. Bernt Schiele, examinerProf. Dr. Paul Lukowicz, co-ethDate of Submission: 26 of May, 2009thDate of Defense: 7 of July, 2009Darmstadt, 2009D17AbstractIn this thesis we use algorithms on data from body-worn sensors to detect physical ges-tures and activities. While gesture recognition is a promising and upcoming alternative toexplicitly interact with computers in a mobile setting, the user’s activity is considered animportant part of his/her context which can help computer applications adapt automati-cally to the user’s situation. Numerous context-aware applications can be found rangingfrom industrial to medical to educational domains. A particular emphasis of this thesis isthe recognition of short activities or quick actions, which often occur amid large quantitiesof irrelevant data.Embedded in different application scenarios, we focus on four challenges in gestureand activity recognition: multiple types and diversity of activities, high variance in perfor-mance and user independence, continuous data stream with large background and finallyactivity recognition on different levels. We make several contributions to overcome thesechallenges.

Sujets

• Informatik

Informations

• Publié par: technischen_universitat_darmstadt
• Publié le: 01 janvier 2009
• Nombre de lectures: 41
• Langue: English
• Poids de l'ouvrage: 7 Mo

Extrait