finding gait cycle estimation

Hi

I have gait verification project that contain gait cycle estimation as main step. Gait cycle estimation is computed by using the aspect ratio of the silhouette bounding box to obtain the beginning and ending frames for a complete gait cycle. This step contain 5 sub step:

• Aspect ratio of silhouette bounding box
• Normalization of aspect ration followed by mean filtering.
• autocorrelation signals
• first-order derivative signals of autocorrelations
• Peak positions indicating the periods.

I am try to apply it by using matlab R2013b and this is my code:

clear 
close all

    clc

    input_dir = 'C:\Users\HP\Desktop\gait\Mobo\moboBgSub\04022\slowWalk\vr17_7';
    filenames = dir(fullfile(input_dir, '*.pbm'));
    tot_ims1 = numel(filenames);

    for n = 1:tot_ims1
    filename = fullfile(input_dir, filenames(n).name);
    img1_p1 = imread(filename);
    img1_p1 = bwareaopen(img1_p1,50); 
    se = strel('square',5);
    s_im= imdilate(img1_p1, se);
    s_im= imclose(s_im, se);

    % Label the 2 levels thresholded image
    [LabeledImage, Num]=bwlabel(s_im);

    % Calculate the area of different components in LabeledImage2
    STATS = regionprops(s_im,  'Area');

    % Get the label of the biggest area in LabeledImage. This will be the human
    [MaxArea,MaxAreaLabel]=max([STATS(:).Area]);

    % Binary detected breast region
    BinaryRegion=LabeledImage==MaxAreaLabel;

    labeledImage = bwconncomp(BinaryRegion,8);
    measurements = regionprops(labeledImage,'BoundingBox','Centroid');
    bb = measurements.BoundingBox;
    bco = measurements.Centroid;
    AR(n)= bb(4)/bb(3);

    end
    fgs=5;
    subplot( fgs,1,  1)
    plot(AR, '-o')
    title('aspect ratio')

    %Normalization of aspect ration followed by mean filtering
    subplot( fgs,1, 2)
    mean_sub= AR- mean(AR);
    std_div= mean_sub/std(mean_sub);
    smooth_wave= smooth(std_div, 17, 'moving');
    plot( smooth_wave )
    title('smoothed')

    % * autocorrelation signals
    subplot(fgs,1, 3)
    autoc= xcorr(smooth_wave);
    plot( autoc )
    title('autoc')

    % * first-order derivative signals of autocorrelations
    % * Peak positions indicating the periods.
    subplot(fgs,1, 4)
    df= diff(autoc);
    plot( autoc )
    [PKS, LOCS]= findpeaks(df);
    hold on
    plot(LOCS, PKS, 'ro')
    title('deriv')

    Distance_Peaks = (max(diff(LOCS))+ mean(diff(LOCS)))/2;
    gait_period= round(Distance_Peaks)

this is the output:

It not given the peak correctly and the value of gait cycle is not logic

How to solve this problem?

Thank you

Answer 1

[Image Analyst]

I don't understand the question. You asked it to find the part of the curve with the steepest slopes, and it did, and it plotted a red circle over the steepest slope part of the curve. So that part is right. What is the problem? Perhaps if you also plotted the slope curve itself, "df", it would help you realize that the circles are really at the peak of the derivative????

And I don't understand "the value of gait cycle is not logic" <== what does that mean?