Human Activity Recognition using Motion History Algorithm

In this research, I have worked to recognize various human actions and activities using Motion History Algorithm. Firstly, I studied different techniques already implemented by various researchers in this field like motion image energy etc and finally develop an approach. In this approach, videos are converted into image frames and the images are pre-processed plus background subtraction is computed for these frames. Then, the motion history of images is computed, and features are extracted. After that, Discrete Cosine Transform (DCT) of the Region of Interest is computed which is then thresholded for classification. The proposed algorithm is able to distinguish between various human actions sittings, standing, hand waving etc. 1.1. Overview: Since the demands of the applications are increasing on daily basis therefore automatically recognition of human activities is in our sights. In surveillance environment, the detection of not normal activities will be very helpful in production of alerts for potential crime. In entertainment environment, this recognition will update the human computer interaction (HCI). In healthcare system, recognition of activities will provide help the rehabilitation of patients. Generally, recognition of human activities is divided into two levels of representations as shown in figure 1: 1. Low-level core technology. 2. High-level applications. In core technology, there are 3 main stages for processing any activity. These stages are object segmentation, feature extraction and classification algorithms. First of all human object in a video sequence will be segmented out. The main characteristics are carefully extracted and the properly represented by labels or features such as: • Shape. • Silhouette. • Colors. • Poses. • Body motions. So in a nut shell, once the features are extracted, the algorithms are applied for the detection, recognition and classification of human activities. High level of applications only works on the results which are obtained when the algorithm are successfully applied on surveillance environments, entertainment environments or healthcare systems.