A modeling method for human actions considering their temporal and spatial differences

Recently, systems to support or recognize human actions have been desired, which must have a certain human action model. Based on this reason, a new modeling method of human actions is proposed in this paper. In the proposed method, a human action model is statistically generated by extraction from enormous data on human actions obtained by long-term monitoring with sensors. Therefore, human actions can be modeled without previous knowledge. In addition, a human action model is generated considering the significance of the differences of not only spatial patterns but also temporal ones on human actions. In the proposed method, a Hidden Markov Model (HMM) and a Hidden Semi Markov Model (HSMM) are properly used to express a human action pattern depending on applications. This is because the amount of the significance of the temporal difference is different depending on the purpose of the use of the human action model. In addition, the relationship between the situation and a human action is expressed by a If-Then-Rule style explicitly. Therefore, an obtained human action model has high readability.