If you remember back to high school health class, heart rate is controlled unconsciously by the autonomic nervous system (ANS). Within the autonomic nervous system there are the sympathetic (SN) and parasympathetic nervous systems (PN). When an exciting situation is presented, the SN kicks in with the famous “fight or flight response,” which prepares the body for quick mobilization. Afterward, the PN sets in and acts as a dampener to allow the body to resume activities it would conduct during rest1.
Instead of seeing heart rate as a one-way street (excitation or no excitation), it is more of a two-way street of excitation and calming in order to prepare for the situations at hand. With that being said, we can learn quite a lot psychologically from the lowering of heart rate. Lower heart rate coincides with higher attention levels, as just seen in the bear example where the hiker was initially more focused2,3.
In the lab, we observe this by means of an EKG (electrocardiogram). Electrodes placed on the skin can pick up the changes in electrical pulses from the heart. This signal is run through an amplifier and a digital converter to the recording software. From there it is filtered and analyzed for beats per minute. Other than counting beats while staring at a watch, this is the most basic and time-tested way of finding heart rate, the gold standard method. One common alternative is PPG (photoplethysmogram), which involves illuminating the skin and measuring light absorption, which changes as blood pumps through capillaries.
*****
The two methods described above require special electronic equipment and physical contact with the subject. But what if you want to measure heart rate without any of that? Can it be done remotely? What if you wanted to find heart rate by just looking at the person (or rather, looking at a video of them)? A team at MIT’s Computer Science and Artificial Intelligence Laboratory believes this is possible. Using a process they developed called Eulerian Video Magnification [http://people.csail.mit.edu/mrub/vidmag/], they created a means to detect seemingly unnoticeable movement in digital video.
First, a person sits in front of a video camera (any camera, which may be a low-quality webcam), and through the software, they select a region on their forehead that they want to analyze. Next, the software breaks down the region into three hues: red, green, and blue. It keeps only the green hues because they show the greatest change in color value as blood circulates through the forehead. Finally, it measures the change in color value across all the pixels of the region and their relationship to each other. Through this and a lot of complicated calculation, it finds a pattern: heart beats.
*****
http://www.huffingtonpost.com/2012/12/24/affectiva-emotion-recognition-technology_n_2360136.html), also uses this heart rate detection technology in their analysis. This software has been around for over two years, so it makes sense that it had made the move from experimentation to application. Nonetheless, I was curious: just how accurate is this webcam heart rate, and should the data obtained be considered on par with data gained by traditional means (EKG)?
I was able to obtain the webcam heart rate software called “Pulse Capture” for free from a file-sharing site [http://sourceforge.net/projects/pulsecapture/]. There seem to be multiple versions of the software because the code is open-source, but I picked one of the more popular options. For my basic, informal experiment, I decided to record 5 minutes of sitting still, absolutely calm. I hooked myself up to the EKG and fired up Pulse Capture. Here’s what I got:
As you can see, it’s not very pretty. I ran some numbers, and the webcam heart rate is off by an average of 2.5 beats per minute throughout the entire session. It’s off by 23 BPM at the point where it strays the most from EKG. That’s 4% off on average and 32% off at its max. I don’t know about you, but personally I’m not a big fan of those numbers.
Looking at this data, I see places where errors in acquisition occurred, namely at seconds 1, 65, 145, and 185. The issue with this software is that it needs to “lock on” to the region of interest first before it calculates heart rate, and any movement, no matter how slight, disrupts the process. The amount of stillness I had to maintain to keep it locked on was very unrealistic given all the fidgeting a person is prone to do during the course of looking at media on a screen.
*****
THE VERDICT:
My verdict on this webcam heart rate technology is that it is not properly adapted quite yet to be used for real-time, phasic responses to time-dependent variables. In non-scientist jargon, it’s not reliable enough to see if heart rate changes in reaction to anything specific. Though I believe it could be of some use for more general, tonic applications, i.e. before and after a workout or anything with more global and obvious differences. It just isn’t reliable enough to generate a steady flow of accurate numbers throughout the entire course of a session.
References:
1http://en.wikipedia.org/wiki/Autonomic_nervous_system
2Porges, S.W. and Raskin, D.C. “Respiratory and heart rate components of attention.” Journal of Experimental Psychology 81.3 (1969): 497-503. Print.
3Coles, M.G. “Cardiac and respiratory activity during visual search.” Journal of Experimental Psychology 96.2 (1972): 371-379. Print.
