Proposed Method Based on Pose Estimation for Complex Activities

in this paper we propose a new algorithm for complex activities our approach accomplished all the requirements of GPLVM. It is suitable for both learning and inference of probabilistic models. The experimental performance has been measured on the basis of Benchmark data. Our propose methodology effectively learn latent spaces of complex multi activities data sets in a computational efficient manner. In addition we also introduce a new procedure for learning latent spaces incrementally. Our proposed model cannot tackle new training set without relearning. Our new method is very flexible and easily applied online settings without extensively repeating relearning method. It has positive impact in applications such as robotics, where domain adaptation plays a vital role for accurate prediction features.