A Classifier Approach using Deep Learning for Human Activity Recognition

Research in the field of Human Activity has been motivated by a number of factors. Whether it is to derive human based activity behavior for 3D models, or for identity recognition using gait, or for medical purposes, HAR is indispensable. Our research proposal a generic framework for recognition of various human activities namely-Jogging, Walking, Sitting, Standing, Upstairs, Downstairs. We applied a number of standard machine learning algorithms and have got the best results by using the Random Forest algorithm approach. Human Activity Recognition is an important aspect of computer vision and its applications. It is highly significant, as it gives us an automated analysis of occurring events and their context. Our model gives us the best accuracy of 98%.