Human Movement Detection using Recurrent Convolutional Neural Networks
M. Kusuma1, M. Vijaya Lakshmi2, K. Sai Krishna3
1M. Kusuma Sri, Assistant Professor, Anurag Group of Institutions Hyderabad (Telangana), India.
2M. Vijaya Lakshmi, Assistant Professor, AGI, Hyderabad (Telangana), India.
3K. Sai Krishna, Assistant Professor, AGI, Hyderabad (Telangana), India.
Manuscript received on 06 October 2019 | Revised Manuscript received on 20 October 2019 | Manuscript Published on 26 December 2019 | PP: 348-351 | Volume-8 Issue-12S October 2019 | Retrieval Number: L109010812S19/2019©BEIESP | DOI: 10.35940/ijitee.L1090.10812S19
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Human Movement detection is vital in Tele-presence Robots, Animations, Games and Robotic movements. By using Traditional methods with the help of sensor suits it is difficult to find and interpret the movements. As it includes so much sensor data which is difficult to interpret, find the action and send to long distances. It is also very expensive and bulky too. Image processing and computer vision provides a solution to detect and interpret Human movement based on R-CNN approach. It is cheap, easy and light weight algorithm. It takes the video input and divides it in to frames, then it is Human body is separated for the background image. This paper mainly focused on skeleton, its major points and its relative positions in successive picture frames. A set of frames (Video) is given as input to the model, so that the model compares the coordinates of the successive frames and estimates the movement. First, the human is identified and separated from the rest of the image by drawing a bounding box around the human by using CNN (Convolution neural networks), then by applying R-CNN human is segmented and converted to skeleton. From the shape of the skeleton we can identify whether the skeleton is that of a human or not. Comparing the relative coordinates of skeletons extracted from frames photographed over time gives the movement of the human and its direction.
Keywords: Video, Deep Learning, R-CNN.
Scope of the Article: Neural Information Processing