Recognition of abnormal human activity using the changes in orientation of silhouette in key frames

In this paper, a simple and robust method has been presented to detect various abnormal activities that can be easily used for elderly monitoring application. The monitoring of elderly persons has become an essential concept in today's world which is moving at a rapid pace for the safety of elders. Due to the fact that some abnormal activities have adverse effects on physical and mental conditions of oldsters, the proposed method monitors the human activities from the video feed and generates appropriate response to the situation. The objective of this monitoring is to detect such abnormal activities. Changes in the orientation of the approximated ellipse around the human body (silhouettes) are used for detection of different abnormal human behaviour. Extracted key features obtained are then passed to a K-NN classifier for categorizing the fall events or other abnormal activity. The high recognition rate obtained from the experimental results highlights the acceptable performance of this proposed system to detect abnormal activities and the implementation of the system for real time monitoring.