Hierarchical recognition of daily human actions based on continuous Hidden Markov Models

Abstract

This paper presents a recognition method of human daily-life action. The method utilizes hierarchical structure of actions and describes it as tree. We modelize actions by Continuous Hidden Markov Models which output timeseries feature vectors extracted by Feature Extraction Filter based on knowledge of human. In this method, recognition starts from the root, competes the likelihoods of childnodes, chooses the maximum one as recognition result of the level, and goes to deeper level. The advantages of hierarchical recognition are:1)recognition of various levels of abstraction, 2)simplification of low-level models, 3)response to novel data by decreasing degree of details. Experimental result shows that the method is able to recognize some basic human actions.

Survey on classifying human actions through visual sensors

“…Other forms of HMMs have been developed to handle more specific problems associated with HMM based action recognition systems (Liu and Chua 2006;Wilson and Bobick 1999;Oliver et al 2000Oliver et al , 2002Gong 2004, 2006;Gong and Xiang 2003;Antonakaki et al 2009;Kawanaka et al 2006;Chakraborty et al 2008;Herzog and Kruger 2009;Mori et al 2004;Brand et al 1997;Del Rose et al 2011;Zhang et al 2006;Fin et al 1998;Natarajan and Nevatia 2007;Herzog et al 2008). Wilson and Bobick (1999) use a Parametric Hidden Markov Model (PHMM) to recognize gestures.…”

Searching Video for Complex Activities with Finite State Models

“…There has been a great deal of interest in models obtained by modifying the HMM structure, to improve the expressive power of the model without complicating the processes of learning or inference. Methods include: coupled HMM's ( [10]; to classify T'ai Chi moves); layered HMM's ( [22]; to represent office activity); hierachies ( [21]; to recognize everyday gesture); HMM's with a global free parameter ( [29]; to model gestures); and entropic HMM's ( [9]; for video puppetry). Building variant HMM's is a way to simplify learning the state transition process from data (if the state space is large, the number of parameters is a problem).…”

Searching for Complex Human Activities with No Visual Examples

“…There has been a great deal of interest in models obtained by modifying the HMM structure, to improve the expressive power of the model without complicating the processes of learning or inference. Methods include: coupled HMM's (Brand et al 1997; to classify T'ai Chi moves); layered HMM's (Oliver et al 2004; to represent office activity); hierarchies (Mori et al 2004; to recognize everyday gesture); HMM's with a global free parameter (Wilson and Bobick 1999; to model gestures); and entropic HMM's (Brand and Kettnaker 2000;for video puppetry). Building variant HMM's is a way to simplify learning the state transition process from data (if the state space is large, the number of parameters is a problem).…”