Human Activity Recognition System.

Date of Submission

December 2010

Date of Award

Winter 12-12-2011

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science


Electronics and Communication Sciences Unit (ECSU-Kolkata)


Mukherjee, Dipti Prasad (ECSU-Kolkata; ISI)

Abstract (Summary of the Work)

The main objective of this project is to identify human activities mainly running and walking. In this report, we present a method for human activity Recognition in video. Human activity recognition from video streams has applications in choreography, sports, security surveillance, content based retrieval motion analysis, virtual reality interfaces, robot navigation and recognition, video indexing, browsing, HCI etc. The complexity involved in human activity varies from simple hand gestures to many body parts a lot.We build an approach to analyze the periodicity in human actions. Here we have considered only two types of activities. They are running and walking. Our method exploits the correlation between the frames for 3 seconds length of time, then identifies the activity periodicityThe system consists of following stages:i) Trackingii) Feature Extractioniii) Classification.Of the above stages, before we track an object, first we are manually selecting the region of interest by clicking on a pixel. We consider a region of M X N sized window around the pixel of interest making pixel of interest as the center of the window. We find the best match of the region in next frame and we track an object till all the frames, will give the trajectory of each and every bounding box that contains our pixel of interest. In this report we mainly presented an approach using the waveforms of a tracking pixel i.e. trajectory of every pixel of interest and finding out their properties using signal processing techniques, extracting features from them and train the system with support vector machines and classify the new videos.Several approaches for activity recognition have been reported in the literature [l]. Previous approaches employed methods such as time-delay neural networks [5], Hidden Markov Models [3][4] or dynamic time warping [2] to recognize hand gestures and articulated human activity.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


This document is currently not available here.