In this paper, we propose a novel framework for semantic medical event characterization and detection by using principal video shots and semantic principal video shot classification. Specifically, the framework includes: (a) A semantic medical event characterization technique by using principal video shots in a specific surgery education video domain. (b) An automatic principal video shot detection algorithm by determining the domain-dependent and event-driven salient objects. (c) A semantic medical event detection technique by using Bayesian classifier, where the classifier parameters and structure are determined automatically by an adaptive Expectation-Maximization (EM) algorithm. For semantic medical event detection in a specific surgery education video domain, our technique achieves overall \approx 87:3% accuracy for four pre-defined semantic medical events.
Principal Video Shot: Linking Low-Level Perceptional Features to Semantic Video Events
2003-06-01
425138 byte
Conference paper
Electronic Resource
English
British Library Online Contents | 2004
|Object Level Frame Comparison for Video Shot Detection
IEEE | 2005
|Asking the user: a perceptional approach for bicycle infrastructure design
Taylor & Francis Verlag | 2022
|Video Indexing Through Integration of Syntactic and Semantic Features
British Library Conference Proceedings | 1996
|Study on human perceptional JND of pedal force from an active accelerator pedal
British Library Conference Proceedings | 2005
|