Non-Contact Automated Heart Rate Measurements using Video Imaging

Sep 25, 2015
3 min read

Phenomena like heart rate (HR), heart rate variability (HRV), respiratory rate (RR), blood pulse pressure, blood glucose concentration and oxy hemoglobin saturation are vital for asserting the physiological state of an individual. Conventional methods used to capture them give good signals but are invasive, expensive and are not portable. The flow of blood below the skin has certain parameters like blood volume, blood velocity, blood pressure and blood flow rate which change over time. These cause changes in the spectra of light transmitted through or reflected form the skin. When captured using a video camera and analyzed after processing, this light information reveals certain health parameters which include HR, HRV, RR and many others.

Human heart rate is one of the most important and commonly measured parameter used to know the medical condition of an individual. It is vital to the proper functioning of heart. Heart rate data can be used to indicate several parameters including, the presence of disease, presence of stress or fatigue and even to know if there are any blockages in the arteries of the heart. The current methods for monitoring these parameters are invasive, expansive and not portable, most of them typically require patients to strap sticky electrodes, bulky sensors and chest straps and on their bodies. These are not only irritating but also can leave severe marks on the skin if not well taken care of.

We have used a new method of caliculating the heart beat using Independent Component Analysis and using the concept of photo-plethysmography(PPG). PPG is based on the concept of measuring changes in reflection of light because of volumetric change in concentration of blood. Experimental results show that the algorithm produces heart rate very close to that of the observed values. We have also explored the Eulerian Video Magnification method, which reveals subtle changes in motion and color during during respiration, artery pulse motion and many other invisible motions.Most of these motions are not visible to human eye.In this projected we narrow our attention to change in human skin color during respiration which can help in finding heart beat and respiratory rate.

The basic approach is, given a video sequence as an input, we analyze the color variation in each pixel over time and amplify the variation and add back to the original video which results in a magnified version of the video.

HEART RATE APPROXIMATION USING ICA APPROACH

Independent Component Analysis is a type of blind source separation method where we do not know the source signal and we try to find the sources responsible from a large number of observed signals that are composed of linear mixtures of the underlying sources.

In our case, the underlying source signal is the cardiac pulse which modifies the reflection of light due to volumetric changes in the facial blood vessels. These changes are picked up by the camera and are recorded and stored in a video. The red, green and blue sensors of the camera, pick up a mixture of PPG signal along with other sources of fluctuations such as noise and motion artifacts. Each sensor picks up a mixture of original source signal with slightly different weights.

Experimental Procedure:

All measurements were performed with minimal light conditions. They were performed indoor with sunlight as only source of illumination. All videos were recorded using the front camera of a mobile(iPhone 4s).All the videos were recorded at 29 frames per second in 24-bit at a resolution of 640x480 pixels. Each video was taken for a duration of 10 seconds. So, each video recorded consisted of 240 frames. All videos were saved in MOV format and were used as the input to the matlab for further processing. Cardiio, an application developed based on the research in MIT Media labs, has been used to compare the results. Cardiio is a very popular application available on app store, which calculates the heart rate based on principle of light reflection using the concept of photoplethysmographic(PPG) signal.