Files in this item

FilesDescriptionFormat

application/pdf

application/pdfWang.pdf (294kB)
(no description provided)PDF

Description

Title:Reading Physiological Signals from Faces
Author(s):Wang, Le
Contributor(s):Ahuja, Narendra
Subject(s):Honorable Mention Undergraduate Image of Research Contest 2014
Video processing
Physiological signal
Electrical and Computer Engineering
Abstract:My research is on non-contact heart rate detection using temporal analysis of face video captured by ubiquitous RGB webcams. The cyclical movement of blood causes the color variations on the face, and the pulse via abdominal aorta andthe carotid arteries. Heart rate is a critical vital sign of physical condition in medical diagnosis. There are also emergingneeds for non-contact, low-term and accessible cardiac pulse estimation with wide applications in health monitoring,emotion assessment, and human computer interaction. To investigate the relationship between temporal color signals on the face regions, I use cross-correlation to measure thesimilarity between color variations on different face regions. The image on the left is the cross-correlation coefficient mapwith respect to the signal extracted at the forehead. The image on the right is the overlaid result of a single face videoframe and the correlation coefficient map. We temporally band-pass filter each signal to reduce noises. From the resultantmap, we could see high similarity between color signals on skin regions and the spatial pattern emerged as caused by the underlying blood flow. Our main research goal is to develop robust heart rate detection in unconstrained environments. Awarded honorable mention in the Undergraduate Image of Research Contest 2014. For more information about the Image of Research--Undergraduate Edition go to: http://go.library.illinois.edu/imageofresearch_uredition
Issue Date:2014-05
Type:Text
Image
URI:http://hdl.handle.net/2142/49134
Rights Information:Copyright 2014 Le Wang
Date Available in IDEALS:2014-05-15


This item appears in the following Collection(s)

Item Statistics