Reducing Spurious Activity Transitions in a Sequence of Movement
Joûef Stefan · 2009
Activity recognition is a fundamental task for anal yzing human behavior and it is usually achieved with a cl assifier. Although the classifier might be very accur ate, it may still produce some false detections and consequ ently, spurious state transitions ‐ the transitions that do not occur in reality. This paper examines two approache s for reducing spurious activity transitions. The fir st approach is based on cost-augment grammar classification namely Sequential Grammar-based Classifier, while the second approach uses hidden Markov models . The paper outlines a basic theoretical background o f the methods and describes the implementation procedure. The results showed that both methods successf ully reduced spurious transitions and improved classific ation accuracy.