%hr = heartRateEstimation(imgDirectory, fps, windowDuration, windowShift)
%%%%%%%%%%%%%%%%%
% hr = heartRateEstimation(imgDirectory, fps, windowDuration, windowShift)
%
% Task:
%
% Inputs:
%	-imgDirectory
% 	-fps
%	-windowDuration
%	-windowShift
%
% Output:
%	-hr:
%
% Author: Thomas Périn
% Date: 17/02/2023
%
% Note: images were extracted with ffmpeg -i "video.mkv" ../img/gg_%04d.png
%%%%%%%%%%%%%%%%%

clc
clear all

%temp var
%imgDirectory = 'img';
fps = 15;
windowDuration = 30;
windowShift = 1;
roi = [280 580 600 850];

% global
redChannel = [];
greenChannel = [];
blueChannel = [];

%list the images
%listImg = dir([imgDirectory '/*.png']);
listImg = dir(['*.png']);

for l_img=1:windowDuration*fps
	% load the current image
	%image_original = imread([imgDirectory '/' listImg(l_img).name]);
  image_original = imread([listImg(l_img).name]);
	%image(image_original);
  %imshow([listImg(l_img).name])

	% crop the current image around the face
	image_face = image_original(roi(1):roi(2), roi(3):roi(4), :);
	%image(image_face);

	% spatial average for r, g, b channels
	r_mean = mean(mean(image_face(:,:,1)));
	g_mean = mean(mean(image_face(:,:,2)));
	b_mean = mean(mean(image_face(:,:,3)));

	% store the current "average" pixel in global arrays
	redChannel = [redChannel r_mean];
	greenChannel = [greenChannel g_mean];
	blueChannel = [blueChannel b_mean];
end

% estimate temporal average and standard deviation
redChannel_avg = mean(redChannel);
redChannel_std = std(redChannel);
greenChannel_avg = mean(greenChannel);
greenChannel_std = std(greenChannel);
blueChannel_avg = mean(blueChannel);
blueChannel_std = std(blueChannel);

% normalize your data
%first define length of our data
L = columns(greenChannel);
for i=1:L
  greenChannel_normalized(i) = (greenChannel(i) - greenChannel_avg)/greenChannel_std;
end

greenChannel_fft = fft(greenChannel_normalized);

% power spectrum (https://www.mathworks.com/help/matlab/ref/fft.html)
P2 = abs(greenChannel_fft/L);
P1 = P2(1:L/2+1);
P1(2:end-1) = 2*P1(2:end-1);
f = 15*(0:(L/2))/L;
plot(f,P1)

% find the peak in the range ([0.75 4] Hz)
% max -> value, index

low_index = 0.75*L/15+0.5
high_index = 2*L/15
a = max(P1(low_index:high_index))
heart_rate = f(a)


% convert the index from Hz to bpm

% determine heart rate