Gait variabiliy analysis for CLBP dataset; 1) determine location of turns, 2) create epochs of 1 minute for each participant, 3) calculate the paramters for each minute and overall. Pre-processing controls not the same, script will be pushed later on.

DetermineLocationTurnsFunc.m

@@ -0,0 +1,68 @@
+function [locsTurns,FilteredData] = DetermineLocationTurnsFunc(inputData,FS,ApplyRealignment,plotit,Distance)
+% Description: Determine the location of turns, plot for visual inspection
+
+% Input: Acc Data (not yet realigned)
+
+%Realign sensor data to VT-ML-AP frame
+
+if ApplyRealignment % apply relignment as described in Rispens S, Pijnappels M, van Schooten K, Beek PJ, Daffertshofer A, van Die?n JH (2014).
+    data = inputData(:,[3,2,4]); % reorder data to 1 = V; 2 = ML; 3 = AP
+    % Consistency of gait characteristics as determined from acceleration data collected at different trunk locations. Gait Posture 2014;40(1):187-92.
+    [RealignedAcc, ~] = RealignSensorSignalHRAmp(data, FS);
+    dataAcc = RealignedAcc;
+end
+% Filter data
+[B,A] = butter(2,3/(FS/2),'low'); % Filters data very strongly which is needed to determine turns correctly
+dataStepDetection = filtfilt(B,A,dataAcc);
+% Determine steps
+% Explanation of method: https://nl.mathworks.com/help/supportpkg/beagleboneblue/ref/counting-steps-using-beagleboneblue-hardware-example.html
+% From website: To convert the XYZ acceleration vectors at each point in time into scalar values,
+% calculate the magnitude of each vector. This way, you can detect large changes in overall acceleration,
+% such as steps taken while walking, regardless of device orientation.
+magfilt = sqrt(sum((dataStepDetection(:,1).^2) + (dataStepDetection(:,2).^2) + (dataStepDetection(:,3).^2), 2));
+magNoGfilt = magfilt - mean(magfilt);
+minPeakHeight2 = 1.2*std(magNoGfilt);                                   % based on visual inspection, parameter tuning was performed on standard deviation from MInPeak (used to be 1 SD)
+[pks, locs] = findpeaks(magNoGfilt, 'MINPEAKHEIGHT', minPeakHeight2);   % for step detection
+numStepsOption2_filt = numel(pks);                                      % counts number of steps;
+
+diffLocs = diff(locs); % calculates difference in step location
+avg_diffLocs = mean(diffLocs); % average distance between steps
+std_diffLocs = std(diffLocs); % standard deviation of distance between steps
+
+figure;
+findpeaks(diffLocs, 'MINPEAKHEIGHT', avg_diffLocs, 'MINPEAKDISTANCE',9);                                    % these values have been chosen based on visual inspection of the signal
+line([1 length(diffLocs)],[avg_diffLocs avg_diffLocs])
+[pks_diffLocs, locs_diffLocs] = findpeaks(diffLocs, 'MINPEAKHEIGHT', avg_diffLocs,'MINPEAKDISTANCE',10);    % values were initially 5
+locsTurns =  [locs(locs_diffLocs),  locs(locs_diffLocs+1)];
+
+magNoGfilt_copy = magNoGfilt;
+for k = 1: size(locsTurns,1);
+    magNoGfilt_copy(locsTurns(k,1):locsTurns(k,2)) = NaN;
+end
+
+% Visualising signal;
+if plotit
+    figure;
+    subplot(2,1,1)
+    hold on;
+    plot(magNoGfilt,'b')
+    plot(magNoGfilt_copy, 'r');
+    title('Inside Straight: Filtered data with turns highlighted in blue')
+    line([6000,12000,18000,24000,30000,36000;6000,12000,18000,24000,30000,36000],[-4,-4,-4,-4,-4,-4;4,4,4,4,4,4],'LineWidth',2,'Linestyle','--','color','k')
+%     hold on;
+%     for m = 1:size(locsTurns,1)
+%         plot(locsTurns(m),DataStraight.([char(Participants(i))]).LyapunovPerStrideRosen_ML(:,m),'--gs',...
+%         'LineWidth',2,...
+%         'MarkerSize',10,...
+%         'MarkerEdgeColor','g',...
+%         'MarkerFaceColor',[0.5,0.5,0.5])
+%     end
+    hold off;
+end
+% Check if number of turns * 20 m are making sense based on total
+% distance measured by researcher.
+disp(['Number of turns detected = ' num2str(size(locsTurns,1))])
+disp(['Total distance measured by researcher was = ' num2str(Distance)])
+
+FilteredData = dataAcc;
+end