Deep Belief Network based Machine Learning for Daily Activities Classification

Human activity recognition has been a very active topic in pervasive computing for several years for its important applications in assisted living, healthcare, and security surveillance. Many researchers are finding and representing the details of human body gestures to determine human activity. While simple activities can be easily recognized only by acceleration data, our research has focused on the recognition and understanding the various activities in daily living. In this work, we address this problem by proposing approach theory of deep learning with the Deep belief network. Deep belief network comprises a series of Restricted Boltzmann Machines will be formed by superimposed multiple Restricted Boltzmann Machines and training the model parameters for data reconstruction, feature construction and classification. We tested our approach on PASCAL VOC datasets. The experimental results indicate that our proposed approach offers significant performance improvements with the maximum of 79.8%.