OHAR: Optimized Human Action Recognition Paradigm using Optimized Type 2 Neuro-Fuzzy Classifier

Human activity recognition (HAR) is made to identify actions and goals of persons one or more from the images which contain sequence of actions related on environments and actions. However, different issues and challenges are increased in the applications of human activity recognition for improving detection accuracy with different activities. Hence, Optimized Human Action Recognition Paradigm (OHAR) is developed. In the paper, optimized type 2 fuzzy classifier is designed to classify human actions from the image database. The input video is transformed into multiple region in the initial stage. The collected frames are sent to the pre-processing stage for removing noise from frames. After that, the key frame is selected from the image frames by using the Structural Similarity Index (SSIM). Once key frames are selected, the three feature extraction methods are utilized like Space-Time Interest (STI) points, grid shape feature, and coverage factor. Finally, the proposed classifier is proceeding to human activity recognition with selected features set. Here, an optimized type 2 neuro-fuzzy classifier is used for detecting human action. The proposed classifier is enhanced Rider optimization algorithm (ROA). The presentation of proposed method is evaluated based on statistical computations such as accuracy, sensitivity, specificity, recall, and F_Score.