HBA optimized Efficient CNN in Human Activity Recognition

Abstract

ABSTRACT A HAR can be a significant test in different applications, counting human–computer connectivity. Hence, in this paper, developing an efficient background subtraction and a deep learning model aimed at classifying human actions from the videos. The main motive of the research work can be developed efficient technique for identifying human activities in videos using a Convolutional Neural Network (CNN) and hybrid feature extraction techniques. The projected CNN classifier can be developed and combined with the Honey Badger Algorithm (HBA) and CNN in the process of feature extraction. The projected classifier is utilized to identify human actions such as bending, walking and so on. HBA can be consumed to optimize the weighting parameters which enhance the performance of CNN. The projected approach is validated considering Weizmann and KTH datasets. Additionally, the performance metrics are considered to justify the performance of the projected technique such as specificity, sensitivity and accuracy.