Video summary
The video introduces a webcam-based tool designed to measure heart rate variability using remote photoplethysmography (rPPG) technology. This application analyzes video footage captured by a standard web camera to detect subtle changes in skin color caused by blood flow, specifically extracting the pulse signal from the green color channel. The presenter explains that hemoglobin absorbs green light significantly more than red or blue light, which allows the system to isolate the heartbeat signal effectively. By monitoring fluctuations in blood volume during each cardiac cycle, the tool calculates beats per minute and displays a waveform, though the presenter notes that smartwatches can also have errors and welcomes contributions from users with more accurate chest strap devices for comparison.
A key technical discussion in the video focuses on the role of signal processing filters within the tool, particularly the high-pass filter used to ensure measurement accuracy. The presenter demonstrates that removing this filter results in a noisy signal dominated by a direct current (DC) drift, which obscures the actual pulse peaks and prevents valid calculations. This DC drift is primarily caused by slow movements, such as slight shifts in body position or changes in lighting due to camera auto-exposure adjustments. The high-pass filter functions by blocking these low-frequency disturbances while allowing the faster-moving pulse signal to pass through, thereby isolating the relevant physiological data from environmental noise and ensuring a clean waveform for analysis.
The video concludes with an interactive segment where the presenter answers questions about the underlying physics of rPPG and the specific functions of the tool's components. Through this process, it is confirmed that the green channel is selected because hemoglobin's absorption properties make it ideal for detecting blood volume changes, and the high-pass filter is essential for removing slow drifts rather than eliminating high-frequency noise or increasing frame rates. The presentation emphasizes that these tools are part of a broader suite of applications available on their website, inviting viewers to explore further resources and even contribute hardware like chest straps to improve future comparisons. Ultimately, the content serves both as an educational explanation of signal processing in remote heart rate monitoring and as a promotion for the developer's open-source projects.
Read the full video transcript
And all the tools we make are available
on binocular.com. Go check them out.
So first of all, hopefully you know what
this is. If you don't, go check it out.
This tool
is running on the side.
It's measuring your heart rate
variability
through the web camera.
I can make sure there's no hairline.
Control for light.
Don't move for a bit.
It will give you a measurement.
It's that peak there.
You're getting a strong 70 beats per
minute
measurement, and I feel like it's about
right.
Normally I compare it to the smart
watch, but then the smart watch could
also have an error, so
if anyone wants to send me a chest strap
or better device, more accurate device,
you're more than welcome to.
And the question
that we have
is asking what specific characteristic
of blood hemoglobin allows for the
extraction of a pulse signal from a
standard green color channel in the
rppg, remote ppg, out of this tool.
Essentially in the options, let's see if
we
can get it correct cuz we designed this
tool, so we should be able to answer it
correctly.
Hemoglobin
absorbs clear green light more
significantly than red or blue light.
Mhm.
Homeoglobin and the that's um
uh
option A answer.
A option B.
Homeoglobin reflects green light during
peak systolic pressure.
Mhm, not sure actually.
And C, green light green light has the
deepest tissue penetration among
among visible wavelengths.
I don't think it's C.
Reason being
that we
uh
Yeah, we're looking at the reflection.
So, I'm not 100% sure if it's not that
it's not C. Let's read D for a sec. The
green channel
is the only one not affected by
melanin levels in the skin.
Uh this is crazy. I don't actually know.
We have a hint. Consider how the
concentration of blood in skin vessels
affects the light that the camera
perceives.
Concentration of blood in skin levels
affects the light.
Is it A?
Yeah, it's A.
I did kind of guess. I mean, all the
options I kind of knew
it's not C or D. I wasn't sure. Yeah, so
A is correct.
Uh hemoglobin absorbs green light more
significantly than red or blue light.
So, that's why we're extracting the
green channel
out of the video footage.
The fluctuations
in blood volume during a heartbeat cause
measurable changes in green light
absorption, which forms the basis of the
pulse wave.
And that's that's that that wave there.
So, that was a really good question.
Brought to you by
a Gemini model through Notebook LM.
Should we go for a new one?
Did the answering the questions elevate
your heart rate?
If the question's too hard, would your
heart rate increase?
Maybe. Maybe. Probably.
Uh we have another question about the
same tool, this one that you can see on
the screen.
And go check it out as well. Everything
we do
is available on binary chaos.com. This
was the last edition.
This tool is turning your sound coming
through the microphone into brain
activation.
So, go check it out.
Everything else we do is available on
binary chaos.com as well.
There's another a about this tool. The
what is the primary function of the high
pass filter in the real-time signal
amplification microscope, which is this
tool? We can, by the way, uh, change the
high pass, uh, filter
uh, settings.
So, it says here
Yes, if you remove it
you can kind of see the change.
So, we get a noisier noisier signal, do
we?
Yeah, when removing the high pass, and
now we have this DC drift, and we can't
make calculation.
So, the high pass was kind of important.
Yeah, you see, there's no peak there.
That peak at zero indicates a high DC
level.
So, when you enable it, it actually
restarts the thing, so you can't just
see like side by side.
It restarts the whole measurement.
But now you see
that it, uh,
the high cut-off
So, the high pass
is cutting any frequency below 0.7
hertz,
so you don't So, essentially remove the
DC component.
So, let's look at the question and try
and answer it correctly. What is the
primary function of the high pass filter
in this tool? To isolate the dicrotic
notch in the PPG waveform.
Uh,
kind of, but I don't think it's
accurate. To remove slow drifts. That
sounds like it. To remove slow drifts
caused by shadow movements or camera
auto auto exposure. Sure about the
camera auto exposure.
But it's slow movements and things will
generate a lot of DC. So B must be
correct, but let's
Let's see.
Option C, to eliminate high frequency
electronic jitter and sensor noise. So
no, that's what the
That's what the
low pass does.
So it's removing everything above 3
hertz, which we don't really need.
So C is not correct. D option D, to
increase the frame rate of the video
feed. No.
So it must be B.
To So the
high pass filter in this application is
used to remove slow drifts drifts.
Caused by shadow movements, so small
movements or camera auto exposure.
Something with the light.
Uh
They block high pass filters block low
frequency drift while allowing the
faster moving pulse signal to pass
through.
So that's
the answer.
And yeah, all the tools that we make
available on bindingcords.com. Go check
them out.