DEMO_drifter_no_SPM - Example usage of the DRIFTER toolbox without SPM

Description: This file demonstrates the use of the DRIFTER method without SPM. Feel free to play around with the parameters (the defaults are returned by the method). Refer to the toolbox documentation for further details about possible parameters.

See also:

Version: updated 2012-04-17

Copyright: Arno Solin, 2011-2012


Add the toolbox path

  % Add the DRIFTER folder to your MATLAB path (Must be changed!!!)
  addpath /path/to/DRIFTER
Warning: Name is nonexistent or not a directory: /path/to/DRIFTER. 

Make artificial data

  % We simulate very simple artificial data fopr one reference signal,
  % and a one-dimensional brain signal with both white noise and periodic
  % noise.

  % Define parameters of the artificial data
  dt     = 0.1;          % Sampling rate (in seconds)
  T      = 0:dt:100;     % Time instants
  f      = 72 / 60;      % 72 beats per minute, constant frequency

  % Make reference data with fundamental signal + one harmonic
  ref_data  = 1*sin(1*2*pi*f*T) + ...

  % Make true slow brain signal
  brain_data = cumsum(0.05*randn(size(T)));

  % Make simulated brain data (random walk + oscillators)
  true_data =  brain_data + ...
              .2*sin(1*2*pi*f*T) + ...

  % Add noise to true data
  obs_data = true_data + 0.05*randn(size(true_data));

Set up DRIFTER and run it

  % Provide two structures: "data" and "refdata" such that = obs_data;
  data.dt = dt;

  % And for each reference signal
  refdata{1}.dt = dt;
  refdata{1}.freqlist = 60:90; % Vector of possible frequencies in bpm
  refdata{1}.data = ref_data;  % NOTE: If this is left out, the
                               % is used.

  [data,refdata] = drifter(data,refdata);

  % All the parameters are now visible in the structures
  % 'data' and 'refdata'.
DRIFTER version: 20120417           :                        Up to date.
Kalman filtering                    :                              Done.
RTS smoother                        :                              Done.

Visualize result

  % The following code is only for providing a visual interpretation
  % of the estimation results. Three figures are shown:
  % (1) the brain signal (the true signal, a noisy observation of it,
  % the DRIFTER estimate, and the DRIFTER estimate with the measurement
  % noise added back to it), (2) The estimated periodic noise component,
  % (3) The stimated frequency trajectory.

  figure(1); clf
  subplot(2,2,[1 2]); hold on
    plot(T,brain_data,'--k', ...
         T,squeeze(,'-g', ...
         T,squeeze(data.estimate),'-b', ...
    legend('True data','Observed data','Estimate','Estimate+white noise')
    title('\bf Results'); xlabel('Time [s]');
  subplot(223); hold on
    title([refdata{1}.name ' Estimate'])
    xlabel('Time [s]');
  subplot(224); hold on
    ylim([min(refdata{1}.freqlist) max(refdata{1}.freqlist)])
    title([refdata{1}.name ' Frequency'])
    xlabel('Time [s]'); ylabel('Freq (bpm)')