|
Abstract
|
The estimation of the human pose is an interesting computer vision problem. The principles of fuzzy logic can be used to distinguish the human poses due to the existing uncertainty. In this paper, we suggest fuzzy logic-based modeling for identifying human actions. Fuzzy membership functions that highlight the discriminative pose connected to each action are considered for feature extraction. Additionally, to identify human activity, a multilayer perceptron classifier is applied. For a more accurate classification of yoga poses, various methods of estimating poses and identifying key points were discussed in detail. Evaluations of the proposed method on the benchmark datasets indicate the performance of the proposed method.
|