Human Motion Detection Based Upon CNN with Pruning and Edge Detection

Abstract

Human activity detection is basic requirement especially within smart environments like smart homes. The primary requirement of smart environment is energy conservation. To achieve this, first target is to detect the human activities accurately using neural network-based approach. This paper presents a unique combination of CNN with pruning and edge detection mechanism to accurately detect human activities. The entire process of human activity detection is portioned into set of phases. In the first phase, data acquisition is performed. The CNN with pruning and edge detection utilised KTH dataset derived from Kaggle. In the second phase, pre-processing to eliminate the noise from the image frames. In the third phase, edge detection and feature extraction were ensured. In the last phase classification is performed. The result of the human activity detection mechanism is expressed in the form of classification accuracy. High classification accuracy of over 95% is observed.