When it comes to human pose estimation in computer vision, this is regarded as one of the most fundamental and challenging areas. In the realm of analysing human posture, recent advances in machine learning (ML) methods have yielded impressive results. However, researchers are working hard to find ways to improve event classification and adaptive posture estimate. Classifying remote sensing events and performing adaptive posture assessment using our method is novel. To minimise noise and improve classification accuracy, Laplacian of Gaussian filtering (LoG), background subtraction, and body parts detection are used in the pre-processing stages of the procedure. Energy, Cartesian perspective, angular geometric and skeleton zigzag characteristics were then used to the multi-fused data. To improve classification accuracy, we use the Henry gas solubility optimization (HGSO) technique to choose the most important characteristics in our feature vector. SVM and k-NN are examples of expert systems that use ensemble classifiers like k-NN, Random Forest, and RF to evaluate a feature set. Events that span multiple Olympic sports and UT-interaction datasets were divided into categories for our analysis.The RF achieved 83.8 percent of accuracy and 82.3 percent of accuracy, SVM achieved 92.5 percent of accuracy and 91.67 percent of accuracy on these two separate datasets.