A Unified Framework for Human Activity Detection and Recognition for Video Surveillance Using Dezert Smarandache Theory

Trustworthy contextual data human action recognition of remotely monitored person who requires medical care should be generated to avoid hazardous situation and also to provide ubiquitous services in home-based care. It is difficult for numerous. reasons. At first level, the. data. obtained from heterogeneous source have different level of uncertainty. Second level generated.. information can be corrupted due to simultaneous operations. In this paper human action recognition can be done based on two different modality consisting. of fully featured camera and wearable sensor. Computationally event features are got from the images and movement actions are provided by wearable sensor. Human action. realization, we havegitvoenuse both decision and feature level fusion methods are studied by a collaborative classifier.. By using feature levelmethod inputs from different sources are combined before going to classification action. For decision level fusion DsMT is used to combine the outputs from two classifiers, each corresponds any one of the sensor. The proposed frame. works is validated using Berkeley Human action database. Based on this frame work human action recognition can be done effectively with increased level.