Algorithm 1. The complete work of the proposed design.

Input: Action Recognition Datasets

Output: Predicted Action Class

Step 1: Input action datasets

Step 2: Load Pre-trained Deep Models; -Densenet201 -Inception V3

Step 3: Fine Deep Models

Step 4: Trained Deep Models using TL

Step 5: Feature Extraction from Avg Pooling Layers

Step 6: SbE approach for Features Fusion

Step 7: Best Features Selection using Proposed KcWKNN

Step 8: Predict Action Label