This document describes a target heart rate monitor project. It takes a 30 second video of a user's finger and measures color intensity changes to obtain their heart rate in beats per minute. It uses various signal processing methods like brightness computation, band-pass filtering, Fourier transforms, peak detection, and smoothing. These methods extract the heart rate from the signal and produce an EKG graph. The project aims to help users achieve their optimal heart rate for activities by monitoring and advising them in real-time. The document outlines the materials, methods, results, accomplishments, and individual contributions of the three student authors.

1. Target Heart Rate Monitor (THRM)
Susan Hamilton, Omar Nada, and Elijah Willie
skhamilt@sfu.ca, onada@sfu.ca, and ewillie@sfu.ca
Department of Computing Science
CMPT 340
Ghassan Hamarneh

2. I. Abstract
By taking a 30 second video of a user finger and measuring the
change in colour intensities we are able to obtain the heart rate in beats per
minute of the user. This is found by using several methods of computing
the brightness of the signal then passing that under a band-pass filtering.
From this point using a sliding window we in sequence find the fourier
transform of the signal, find the maximum peaks, and smoothing these
peaks. From these sequence of methods we obtained the EKG graph which
provides a visual understanding for the user. Because we already had the
users optimal heart rate of ther user we can advise them on how to obtain
their optimal heart rate for the given activity.
II. Keywords
 Heart Rate or Heart Beat
 Brightness Signal Computation
 Band-Pass Filtering
 Fast Fourier Transform
 Sliding Window
 Sampling Frequency
 Peak Detection
 Smoothing
 Image Noise or Signal-to-Noise Ratio (SNR)
III. Introduction
It is often that we find ourselves wanting to participate in a particular
activity such as studying, exercising, or just waking up or going to sleep,
but complaining about not being in the right state of mind. THRM’s main
purpose is to assist in achieving this optimal state by monitoring the heart
rate. Over a testing period THRM is able to obtain an optimal heart rate
specific to the user. From this point on THRM is able to monitor the users
heart rate using their mobile device and advise in ways to either rase or
lower their heart rate in a short amount of time in order to allow the user to
achieve the optimal state necessary for peak performance in their activity.
The motivation of this project is to assist the user to obtain their own
personalized heart rate monitor. That is to say that not everyone is the same
and what works for one person may not work for another. Also heart rate
could be affected by a person’s height, weight, gender, and general state of
health. THRM is also specialized in that though it will provide initial
activities such as those listed above, it will also allow the user to

3. personalize their own activity tags specific to their needs. No one person is
the same and neither should THRM only cater to a small rage of activities.
In the rest of the report we will cover the material that is used, section
IV, and section V will cover the methods that were used to construct our
results and accomplishments which will then be covered, section VI and
section VII. Following this we will overview the contributions of
individual members, section VIII, as well as acknowledge referenced
material in section, XII. We will also provide our discussions on our
conclusions and what could be coming in the future for this project in
sections IX and X.
IV. Material
 Human subjects
 Mobile device or computer with camera
 30 second clip to obtain the signal required to analyze and
produce a EKG graph
V. Methods
i. Video Signal Analysis
o Since research shows the human heartbeat is
between 60 and 200 beats per minute to avoid
aliasing we followed the Nyquist Sapling
Theorem where we doubled the maximum value.
ii. Brightness Signal Computation
o We looked at the red value and averaged their
values.
o
Where W=frame width
H=frame height
[n,x,y,1] % since looking at red only
video = VideoReader('path to your video file here');
brightness = zeros(1, video.NumberOfFrames);
for i = 1:video.NumberOfFrames,
frame = read(video, i);
redPlane = frame(:, :, 1);
bght(i)=sum(sum(redPlane))/(size(frame,1)*size(frame,2));
end
iii. Band-Pass Filtering
o This step attenuates frequencies not within the
interested region.
o Helps to remove extra noise.

4. o In particular we used second-order Butterworth filter
– applies the ‘maximally flat‘.
BPM_L = 40; % Heart rate lower limit [bpm]
BPM_H = 230; % Heart rate higher limit [bpm]
FILTER_STABILIZATION_TIME = 1; % [seconds]
% Butterworth frequencies must be in [0, 1], 1
corresponds to half the sampling rate
[b, a] = butter(2, [((BPM_L / 60) / v.FrameRate * 2),
((BPM_H/60)/v.FrameRate*2)]);
filtBrightness = filter(b, a, brightness);
% Cut the initial stabilization time
filtBrightness = filtBrightness((v.FrameRate *
FILTER_STABILIZATION_TIME + 1):size(filtBrightness, 2));
iv. Fourier Transform
o We used the discrete fourier trasform to change it
from the time domain to the frequency domain.
o Used fft(signal) in Matlab
o Sliding window was used to repeat peak detection
and smoothing over every .5seconds of a 6 second
frame
o
 Where Fr= frequency resolution
Fs=sampling frequency
N=number of window samples
Tw=window time duration
 Window length effects accuracy: higher the
window time better the Fr but the lower the
time accuracy
v. Peak Detection
o Use findpeaks function in Matlab we find a sample
data if the point is larger than its two neighbours and
then we use max function to find the max peak.
o Then change it to the fourier tranform vector.
% Translate the freq range to indices within FFT vector
rangeOfInterest = ((BPM_L:BPM_H) / 60) *
(size(fftMagnitude, 2) / sampFrequency) + 1;
% Find peaks in the range of interest
[peaksValues,peakIndices]=findpeaks(fftMagnitude(rangeO
fInterest));
% Find the highest peak
[maxPeakValue, maxPeakIndex] = max(peaksValues);
% Translate the peak index to an FFT vector index
bpmFreqIx = rangeOfInterest(peakIndices(maxPeakIndex));

5. % Get the freq in bpm corresponding to the highest peak
bpmPk = (bpmFreqIx - 1) * (sampFrequency /
size(fftMagnitude, 2)) * 60;
vi. Smoothing
o We have now found teh most powerful spot in
frequency
o Now correlate around the fourier transform around
the peak using zero-padding
fftResolution = 1 / WINDOW_SECONDS;
lowFreq = bpmPeak / 60 - 0.5 * fftResolution;
smoothingResolution = SMOOTHING_RESOLUTION / 60;
testFreqs = round(fftResolution / smoothingResolution);
power = zeros(1, testFreqs);
freqs = (0:testFreqs -1)*smoothingResolution + lowFreq;
for k = 1:testFreqs,
re = 0; im = 0;
for j = 0:(size(b, 2) - 1),
phi = 2 * pi * freqs(h) * (j / samplingFreq);
re = re + b(j+1) * cos(phi);
im = im + b(j+1) * sin(phi);
end
%only need to find the maximum, we use power
power(k) = re * re + im * im;
end
[maxPeakValue, maxPeakIndex] = max(power);
smoothedBpm = 60 * freqs(maxPeakIndex);
VI. Results
When we film a 30 second video of a finger placed on the screen
we obtain the heart rate in bpm and a graph showing the heart rate
for that frame and how it changed over time
Sleeping Exercising

6. VII. Accomplishments
For a project such as this, it was difficult to know what was feesable
to accomplish within the given time restrictions as well as without
experience and knowledge in the areas we would be attempting to delve
into. With this being said we were able to accomplish success in many of
the important areas of the proposed project. The fundamental part of the
assignment we put as foremost necessary to accomplish was the functional
heart rate monitor using the camera of a mobile device or laptop. This we
were able to achieve due to online resources (see section XII for our
acknowledgements). With a some tweaking we were able to apply multiple
methods (described in section V) in taking a signal and removing noise to
find the heart rate of the individual. We were also able to find a wealth of
resources on heart rate. We did meet an opstacle here with ensuring that the
information we obtained was FDA approved. With this being said we were
able to find quite a few works that cited information which was FDA
approved. With this problem also came a fundamental lack of knowledge
on heart rate. In using a testing period where the user would have their
heart rate tested when they were in their optimal state we were able to
overcome the need to know in depth knowledge on what science says an
optimal heart rate should be given a wide range of factors such as weight,
height, and gender. Another detail is to know by how much deviation of the
heart rate should be considered over or under the optimal heart rate and we
did do a bit of testing with various activities, but this is an area that could
use further testing and research. The one area that proved to be the largest
obstacle was that of the graphic user interface. Initially we had decided to
focus on an iOS device application using xCode. Though many hours were
spent on this, due to a lack of previous experience by any of the members
this area was progressing rather slowly. We also came to realize that Apple
required a license to publish applications and that we would perhaps not
wish to make an iOS device application, but instead an Android application
as well as a laptop or desktop application. Do to the time constraint we
decided that it was more important to ensure that the actual software was
working and that we obtained a plentiful of information was obtained for
the application before we focussed on producing the GUI.
VIII. Contributions
All members helped one another in the various areas one
advicement or on research and understanding. With this in mind,
below is listed the assigned duties of each member and what their
main focus was on.

7. Omar Nada contributed mainly to the research of all information
pertaining to the heart rate. Information on how to increase and decrease
the heart rate as well as information on general heart rate health.
Susan Hamilton crontributed to the research iOS programming as well
as on programming on other devices such as android and desktop. All GUI
programming was done by Susan. Also did the write up for the project
proposal and the written report as well as produced the oral presentation
poster hard copy.
Elijah Willie contributed to the research and the implementation of
the heart rate monitor in Matlab, focussing on the research of filtering and
noise reduction. Elijah was responsible for ensuring that the software for
producing the heart rate as well as the EKG graph were working. Elijah
also wrote the project update.
IX. Conclusions and Discussions
From the amount of time we had and the understanding we went into
the assignment I would say that what we accomplished and learned was
very good. From the algorithm we used to compute the heart rate and EKG
graph we found that the run time was decent do to many attempts to
simplify computation. Also with an error of approximately 2-3 values of
standard deviations we can declare that the accuracy was very decent. Time
constrants did play a large role in how much we could accomplish which is
a good lesson to learn in order to help us budget our time better in the
future, especially when a GUI must be produced. In general it was a great
way to learn the general application of the methods we learned in class. It
means that something that was initially just theory has way more meaning
to us and we can now broaden what we can apply the methods learned to.
X. Future Work
As mentioned above in section VII, some areas of our project were
not completed fully. One such section is that we would like to obtain more
information on heart rate. Information that we did obtain could be further
inhanced by further research. Also when the application is made we do not
want to provide the same suggestions again and again on how to lower and
raise the heart rate. It would create a more fun and interesting application if
we could provide ideas that were more creative such as meditating for half
an hour, but information such as this is harder to verify as FDA approved
and so this is where research is necessary. Another area that we need to
research is variation in heart rate from activity to activity. Simple tests did
show that the difference in heart rate for running and at rest activities is

8. sufficiently different, but research and testing on variation from studying
and waking up is necessary. On this not we would like to make the
application even more useful by providing a section which includes further
information for the user on what a healthy heart rate is and ways to achive a
better one in order to promote a healthy lifestyle. The last section that
requires future work is continued research into GUIs as well as porting
languages. That is our software code is in Matlab while iOS programming
is in xCode. This is obviously a fundamental future development in order to
actually publish the finished application we will have produced.
XI. References
Wilmore JH, et al. Physiology of Sport and Exercise. 4th ed. Champaign,
Ill.: Human Kinetics; 2008:162.
American College of Sports Medicine. ACSM's Resources for the Personal
Trainer. 3rd edition. Baltimore, Md.: Lippincott Williams & Wilkins; 2009:274.
Sauer WH. Normal sinus rhythm and sinus arrhythmia.
http://www.uptodate.com/index. Accessed June 28, 2012.
Your guide to physical activity and your heart. National Heart, Lung, and
Blood Institute. http://www.nhlbi.nih.gov/health/public/heart/#obesity.
Accessed June 25, 2012.
http://www.mayoclinic.org/diseases-conditions/heart-disease/in-depth/heart-
disease-prevention/art-20046502?pg=2
http://www.heart.org/HEARTORG/Conditions/More/MyHeartandStrokeNews/All-
About-Heart-Rate-Pulse_UCM_438850_Article.jsp
http://www.appcoda.com/enhance-your-simple-table-app-with-property-list/
https://developer.apple.com/library/ios/documentation/iPhone/Conceptual/iPhoneO
SProgrammingGuide/iPhoneAppProgrammingGuide.pdf
XII. Acknowledgements
Ghassan Hamarneh for providing motivation for our project and
Ignacio Mellado for providing detail methodology and code for the success
of the project. http://www.ignaciomellado.es/blog/Measuring-heart-rate-
with-a-smartphone-camera