A Hybrid Deep Neural Networks for Sensor Based Human Activity

Since use of digital camera in daily life increases, more content of video are uploaded and created to the web in the form of large video data format. Act recognition could be a demanded and popular area for research thanks to its potential application in video retrieval, interaction in human-computer and medical. In healthcare, wellbeing, and in the sports monitoring system, it is important to recognize activities performed by the any human body. The recording of the performed activities by human are very important, for acknowledge of wellbeing, identifying training activities for athletes, or knowledge about the activity being carried out by a patient for analysis of pathologies. Action recognition is important and challenging statistic classification task. The main application of human action Recognition varies from Human-Computer Interaction, Content- based video reporting Human fall detection, Visual Surveillance, medium Intelligence, and Video labelling etc. Many researchers have opened human action recognition and each one proposed different solution to clear the issues related this. Generally, they used vision sensors, inertial sensors or combination of them. Machine learning and threshold learning are often used. Deep learning methods are being introduced for better performance and better accuracy. DL is considered as a modern method, for solving human action recognition problems regarding gesture recognition and action recognition. The main application of human action Recognition varies from Human-Computer Interaction, Content- based video reporting Human fall detection, Visual Surveillance, medium Intelligence, and Video labelling etc.