| Summary: | One of the most important aspects of promoting the intelligence of home service robots is to reliably recognize human actions and accurately understand human behaviors and intentions. In the task of action recognition, there are many common ambiguous postures, which affect the recognition accuracy. To improve the reliability of the service provided by home service robots, this article presents a method of probabilistic soft-assignment recognition scheme based on Gaussian mixture models to recognize similar actions. First, we generate a representative posture dictionary based on the standard bag-of-words model; then, a Gaussian mixture model is introduced for the similar poses. Finally, combined with the Naive Bayesian principle, the method of weighted voting is used to recognize the action. The proposed scheme is verified by recognizing four types of daily actions, and the experimental results show its effectiveness.
|
|---|
