load_SensorLogger_directory.m

# Octave implementation to load a directory with SensorLogger files
#
# Expected files: accelerometer, lin_acceleration, light, rotation,
# magnetic_field, gravity, proximity, orientation
#
#
# Usage: this file creates eight global vars (see files_to_load).
# If a = accelerometer, l = lin_acceleration, r = rotation, then:
#
# clf;
# figure('Position', [0, 0, 1700, 800])
# hold on;
# plot(a(:,2), a(:,[3:5]));
# plot(l(:,2), l(:,3),'m', l(:,2), l(:,4),'c',l(:,2), l(:,5),'k');
# plot(r(:,2), r(:,3),'.m', r(:,2), r(:,4),'.c',r(:,2), r(:,5),'.k');
#

1;

global files_to_load;
global loading_directory;

files_to_load = {'accelerometer', 'lin_acceleration', 'light', 'rotation', 'magnetic_field', 'gravity', 'proximity', 'orientation'};

# Create global variables
for i = 1 : length(files_to_load)
  eval( cstrcat('global ', files_to_load{i}));
end

function files = load_directory(directory)
  global files_to_load;
  global loading_directory;



  # Determine of path already has trailing slash
  separator = '';
  if substr(directory, -1, 1) != '/'
    separator = '/';
  end

  loading_directory = strcat(directory, separator);

  for i = 1 : length(files_to_load)
    filename = char(strcat(directory, separator, files_to_load{i}, '.csv'));
    new_data = [];
    if exist(filename, 'file' )
      new_data = load('-ascii', filename);
      # printf('Read %s, size: %i, %i \n', files_to_load{i}, size(new_data));

      # Transform unix nanoseconds timestamp to microseconds offset from start
      new_data(:,1) = (new_data(:,1) - new_data(1,1))/1000;
    end

    # Register global vars in this function
    eval( cstrcat('global ', files_to_load{i}));

    # Assign values to global vars
    eval( strcat(files_to_load{i}, '= new_data;') );
  end
  files = files_to_load;
endfunction



# Get a matrix of all metrics combined.
# The first column is the timestamp. All occurences of all timestamps
# are generated. If a timestamp is missing, the previous values is copies.
# So there is _no interpolation_.
function merged = merge_sensors(save_csv = false)
  global files_to_load;
  global loading_directory;

  for i = 1 : length(files_to_load)
    eval( cstrcat('global ', files_to_load{i}));
    eval( strcat( 'files_', num2str(i), '=', files_to_load{i}, ';'));
  end

  merged = [];
  times = [];
  # get all the first values to init the merged array
  row = [];
  for i = 1 : length(files_to_load)
    is = num2str(i);
    values = eval( strcat( 'files_', is, '(1,3:5);' ) );
    row = [row values];

    time = eval( strcat( 'files_', is, '(1,2);' ) );
    times = [times time];

    eval( strcat('files_', is, '(1,:) = [];'));

  end

  merged = [min(times) row];

  while true

    new_times = [];
    new_row = [];
    sizes = [];
    for i = 1 : length(files_to_load)
      is = num2str(i);

      size_this = eval(strcat('size(files_', is,',1)'));
      sizes(end+1) = size_this;

      if size_this > 0
        time = eval( strcat( 'files_', is, '(1,2);' ) );
      else
        time = smallest_time +1;
      end

      new_times(end+1) = time;
    end

    if length(find(sizes > 0)) == 0
      break;
    end

    smallest_time = min(new_times)
    smallest_indices = find(new_times <= smallest_time);
    non_smallest_indices = find(new_times > smallest_time);

    # Use value of current row for new merged matrix
    # for the values with smallest timestamp
    for i = 1 : length(smallest_indices)
      index = smallest_indices(i);
      indexs = num2str(index);

      row_index_range = [(index-1)*3+1 : (index-1)*3+3];

      if eval(strcat('size(files_', is,',1)')) > 0
        value = eval( strcat('files_', indexs, '(1,3:5);'));
        # remove used values
        eval(strcat('files_', indexs, '(1,:) = [];'));
      else
        value = row(row_index_range);
      end

      new_row(row_index_range) = value;
    end

    # USe value of previous row for new merged matrix
    # for the values with not a smallest timestamp
    for i = 1 : length(non_smallest_indices)
      index = non_smallest_indices(i);
      row_index_range = [(index-1)*3+1 : (index-1)*3+3];
      # value = eval( strcat('row(', num2str(row_index_range), ');'));
      new_row(row_index_range) = row(row_index_range);
    end

    row = new_row;
    merged(end+1,:) = [smallest_time row];

  endwhile

  if save_csv
    csvwrite(strcat(loading_directory, "merged.csv"), merged );
  end

endfunction


# Plot the axis (e.g. 3 for the x) of a specific metric.
# Draw the mean (dashed black line) and standard deviation
# (blue dotted) of a half-overlapping window
# of width samples.
# Optionally use absolute values
# usage: plot_mean_std('lin_acceleration', 3, 80, true )
function plot_mean_std(metric, axis, width = 50, absolute = false )
  global files_to_load;

  for i = 1 : length(files_to_load)
    eval( cstrcat('global ', files_to_load{i}));
  end

  clf;
  hold on;
  eval( strcat('values = ', metric,';' ) );

  # Original values, solid line
  plot(values(:,2), values(:,axis), ':r');

  means = [];
  stds = [];
  times = [];

  for i = width + 1 : width/2 : size(values,1)
    window = values(i-width:i, axis);
    if absolute
      window = abs(window);
    end
    time = values(i,2);
    m = mean(window);
    s = std(window);
    times(end+1) = time;
    means(end+1) = m;
    stds(end+1) = s;
  end

  stairs(times, means, '--k')
  stairs(times, means+stds, ':b')
  stairs(times, means-stds, ':b')

endfunction


function plot_auto_correlation(metric, width = 50 )
  global files_to_load;

  for i = 1 : length(files_to_load)
    eval( cstrcat('global ', files_to_load{i}));
  end

  # clf;
  hold on;
  eval( strcat('values = ', metric,';' ) );

  corr_x = [];
  corr_y = [];
  corr_z = [];
  times = [];

  for i = width + 1 : width/2 : size(values, 1)

    window_first = values(i-width : i - width/2, 3:5);
    window_last = values(i - width/2 : i, 3:5);

    time = values(i,2);



    times(end+1) = time;
    coeff = [coeff; c(1,2) c(1,3) c(2,3)];

  end

endfunction


function plot_corr(metric, width = 50)
  global files_to_load;

  for i = 1 : length(files_to_load)
    eval( cstrcat('global ', files_to_load{i}));
  end

  # clf;
  hold on;
  eval( strcat('values = ', metric,';' ) );

  coeff = [];
  times = [];

  for i = width + 1 : width/2 : size(values,1)
    window = values(i-width:i, 3:5);

    time = values(i,2);
    c = corr(window)
    times(end+1) = time;
    coeff = [coeff; c(1,2) c(1,3) c(2,3)];
  end

  stairs(times, coeff )

  legend('x-y correlation coefficient', 'x-z correlation coefficient', 'y-x correlation coefficient')

endfunction



# Plot the global variables:
# accelerometer, lin_acceleration, rotation, magnetic_field
# accelerometer is default false, because it is the same as lin_acceleration,
# but with the gravity included. Helpful for orientation though.
# magnetic_field is default false, gives orientation to magnetic north.
function plot_values(a = false, l = true, r = true, m = false)
  global files_to_load;

  for i = 1 : length(files_to_load)
    eval( cstrcat('global ', files_to_load{i}));
  end

  clf;
  hold on;

  labels = {};

  if a
    acc = accelerometer;
    plot(acc(:,2), acc(:,[3:5]), '-');
    labels = cat(2, labels, {'X acc', 'Y acc', 'Z acc'});
  end

  if l
    lin = lin_acceleration;
    plot(lin(:,2), lin(:,3), '-m', lin(:,2), lin(:,4),'-c', lin(:,2), lin(:,5),'-k');
    labels = cat(2, labels, {'X lin', 'Y lin', 'Z lin'});
  end

  if r
    r = rotation;

    if (a||l||m)
      printf 'Scaling rotation for visual inspection\n';
      r(:,3:5) = rotation(:,3:5) * 10;
    end

    plot(r(:,2), r(:,3),'.r', r(:,2), r(:,4),'.g', r(:,2), r(:,5),'.b');
    labels = cat(2, labels, {'X rot', 'Y rot', 'Z rot'});
  end

  if m
    m = magnetic_field;
    plot(m(:,2), m(:,3),'.m', m(:,2), m(:,4),'.c', m(:,2), m(:,5),'.k');
    labels = cat(2, labels, {'X mag', 'Y mag', 'Z mag'});
  end

  legend(labels);

endfunction