Video Processing for Health Assessment

The goal of the project was to develop a computer vision algorithm for tracking the heart rate of an individual from a recording of his/her video by computing the Blood-Volume Pulse. To do this, techniques such as Independent Component Analysis, analysis of individual color channels, etc were exploited. Subsequently, using signal-processing techniques this computed heart-rate was used for classifying a person as being psychologically healthy or suffering from anxiety disorders. An overview of the classification pipeline is shown in the figure above.

The project was undertaken under my leadership. The research findings have been published at a top conference.

Publications:

• M. Chatterjee, G. Stratou, S. Scherer, L. P. Morency, "Context-Based Signal Descriptors of Heart-RateVariability for Anxiety Assessment”, IEEE Int’l Conf. on Acoustics Speech and Signal Processing 2014 (IEEE ICASSP 2014).

Project Funding Support:

We are extremely grateful towards the U.S. DARPA for providing funding support to the project.