Abnormal Human Activity Recognition using Scale Invariant Feature Transform

Human detection in camera attract wide interest because of its extensive applications in anomalous incident recognition, counting human being in a crowded area, person classification, and recognition of falling activity for aged people, etc. The paper discuss abnormalities in the human activity and provide efficient solution to detect abnormality. The first step in the proposed work is to capture the video using webcam and then to detect abnormal behavior. The captured video is divided into frames and extract the features such as edges and boundaries using scale invariant feature transform (SIFT). Feature vectors are developed from the extracted features. These feature vectors are compared using Hidden Markov Model with the data set developed to recognize abnormal behavior. If there is a match between feature vectors and available data set, then abnormal human activity is detected and simultaneously an alarm is given for medical assistance. The proposed algorithm is tested on six different activities. The proposed methods achieve accurate recognition.