path: root/theremin.m

clear;

show_graphs = false;

%%%%%%%%%%%%%%%
% "constants" %
%%%%%%%%%%%%%%%

% how many updates per second, determines the length of audio snippets
const_frames_per_second = 100;

const_Fs = 44100;  % sample rate in Hz
const_te = 1/const_frames_per_second; % signal duration in seconds
const_samples_per_frame = ceil(const_Fs * const_te);
const_sample_range = 0:const_samples_per_frame-1;

% the smaller the value, the quicker the fade out
const_fade_speed = 0.975;

% signal "generator" - only gives the input for sin()
%  it does not apply the sin function yet, see below for reasoning
signal_f = @(freq) (freq ./ const_Fs .* 2 .* pi .* const_sample_range);

% dimensions to use for detecting number of hands in frame
zero_hands = size(NaN(0,0));
one_hand   = size(NaN(1,1));
two_hands  = size(NaN(1,2));

deviceWriter = audioDeviceWriter('SampleRate', const_Fs, ...
    'SupportVariableSizeInput', true, 'BufferSize', 3 * const_samples_per_frame);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% init matleap by calling for first frame %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
matleap_frame;

% runtime variables
P = NaN(1000000,3);
count = 1;
done = false;

frequency_pos = 0;
height_pos = 0;

offset = 0;

signal = sin(signal_f(0));

complete_signal = signal;

% time gestures, used for program termination
gesture_count = 0;

% main loop
while gesture_count < const_frames_per_second
    frame = matleap_frame;
    handCount = size(frame.hands);
    
    % slowly decrease volume
    height_pos = height_pos * const_fade_speed;
    
    if isequal(one_hand, handCount)        
        %pos = frame.hands(1).palm.position;
        pos = frame.hands(1).palm.stabilized_position;
        
        frequency_pos = pos(1);
        height_pos = max(150, height_pos);
        
        P(count, 1:3) = pos;
        count = count + 1;
    
        if frame.gesture > 0
            gesture_count = gesture_count + 1;
        else
            gesture_count = 0;
        end
    elseif isequal(two_hands, handCount)
        gesture_count = 0;
        
        %pos = frame.hands(1).palm.position;
        pos = frame.hands(1).palm.stabilized_position;
        
        frequency_pos = pos(1);
        %y_one = pos(2);
        
        P(count, 1:3) = pos;
        count = count + 1;
        
        %pos = frame.hands(2).palm.position;
        pos = frame.hands(2).palm.stabilized_position;

        %x_two = pos(1);
        height_pos = pos(2);
    else
        % no hands, do nothing (or more than 2 (how?! :D))
        gesture_count = 0;
    end
    
    % play current sound
    [signal, offset] = get_theremin_sound_bit(frequency_pos, height_pos, offset, signal_f);
    
    buffer_under_flow = deviceWriter(signal(:)); 
    
    %if buffer_under_flow ~= 0
    %    disp("Buffer ran empty!");
    %end
    
    complete_signal = [complete_signal signal];
end

release(deviceWriter)

%theremin_player = audioplayer(complete_signal, const_Fs);
%play(theremin_player);

if show_graphs == true
    % extract values
    x = P(:,1); % links (-) rechts (+) (LED zu uns)
    y = P(:,2); % oben unten
    z = P(:,3); % vorne (+) hinten (-) (LED zu uns)

    % plot
    figure("Position",[0,0, 1200, 2400]);
    t = tiledlayout(4,1);

    nexttile;
    plot(x);
    ylabel('left right');

    nexttile;
    plot(y);
    ylabel('height');

    nexttile;
    plot(z);
    ylabel('depth');

    nexttile;
    plot3(z,x,y);
    xlabel('depth');
    ylabel('left right');
    zlabel('height');

    %{
    % Plot both audio channels
    N = size(complete_signal,2); % Determine total number of samples in audio file
    figure;
    subplot(1,1,1);
    stem(1:N, complete_signal(1,:));
    title('Audio Channel');
    % Plot the spectrum
    df = const_Fs / N;
    w = (-(N/2):(N/2)-1)*df;
    y = fft(complete_signal(1,:), N) / N; % For normalizing, but not needed for our analysis
    y2 = fftshift(y);
    figure;
    plot(w,abs(y2));
    %}
end

disp('If you want to save your audio, run `audiowrite(<.wav filename>, complete_signal, const_Fs)`');

function [sound,offset] = get_theremin_sound_bit(x, y, offset, generator)
    % the values used here are mostly empirical
    
    volume = y / 1300;
    % How it works:
    %  https://web.physics.ucsb.edu/~lecturedemonstrations/Composer/Pages/60.17.html
    frequency = max(3, (x + 155) * pi); % have at least 3Hz

    % here we generate the array to put into sin(), but offset it with the
    %  the previous frames offset..
    base = generator(frequency) + offset;
    
    % ..which we take from here - doing so stops us from phase jumping
    offset = base(end);
    
    % at last, apply the volume
    sound = sin(base) .* volume;
end