Hidden Markov Model based human activity recognition using shape and optical flow based features

Recognizing human activity is an important area of research in computer vision application. Manual monitoring of all cameras continuously for longer duration is inefficient making auto-detection of activity important. In this paper shape and optical flow features are fused together and used for human activity recognition. Features extracted are found to be efficient as concluded by ANOVA test. Hidden Markov Model are generated for each activity. System is trained and tested in various indoor and outdoor environment. The method adapted is made shape and angle invariant. Accuracy achieved using least square support vector machine classifier is 80% for all activities. Hidden Markov Model resulted in better accuracy as compared to least square support vector machine classifier with accuracy of 100.00% for walking, 100.00% for hand waving, 90% for bending, 84.61% for running and 90% for side gallop activities. 100% accuracy is achieved in recognizing activity in different angle with respect to camera.

I. Introduction

Human activity recognition is one of the ongoing research activity in computer vision field and has a lot of practical application. Detecting activity in real time automatically in outdoor and indoor environment demand robustness of features and classifier. As the distance and angle of person with respect to camera may not be fixed, feature extracted should be rotation invariant. Occlusion is also a major problem in crowded scenario like airport as region of interest extracted may be overlapping making extraction of shape features difficult. Automatic activity recognition is required for detection of abnormal activity in airport, railway station, bus stop and also in shopping mall. Activity recognition can be achieved by differentiating the activities based on it's appearance, shape, interest point, optical flow or motion features [1]. In classifiers, Hidden Markov Model (HMM) and Support vector machine (SVM) classifier are most used classifier for activity classification. Jie Yang et.al [2] proposed blob feature based activity recognition. Seven hue moments were extracted and K nearest neighbour classifier was used for classification. Average accuracy achieved was 95.10%. Hafiz Imtiaza et. al [3] had used spectral domain features for human activity recognition. They had used 2D-discrete Fourier transform for spectral feature extraction and principal component analysis technique for dimension reduction. Authors have claimed to have achieved 100% recognition accuracy on the online Weizmann database. Combined optical flow and shape based features along with support vector machine classifier were tested on online Weizmann dataset by K. G. Manosha Chathuramali et.al [4]. Accuracy claimed by author is 100%. Sabanadesan Umakanthan et.al [5] had proposed binary tree svm based activity recognition. The Histogram Oriented Gradients (HOG) descriptor and the Motion Boundary Histogram were used as feature. Accuracy achieved is 58.2% when tested on online database.