How can I cut the number of frames down?

I have input data from force plates, to count the number of Tappings from a sport test. The measurement must last 15 sec. for an exact and valid result. This is equevanlent to 11250 frames with a frame rate of 750 Hz. Therefor I have to cut the data. I want a solution to do it automatically by define a start and end point with some sort of event, not manually by cutting the graph with uniwait and ginput.

%% Read data

[filename,pathname] = uigetfile('*.mat','Bitte Sprungdaten Mat-Datei auswählen','C:\Users\ifdadmin\Desktop\Local QTM Project\Data\20201007FR02\Running,Squat,CMJ,DJ,Extra\CMJ 2.qtm');

load([pathname, filename]);
% Kraftdaten aus Rohdaten ziehen
Data = struct2cell(load([pathname,filename]));

% Force data 
F1 = (Data{1,1}.Force(1).Force);
F2 = (Data{1,1}.Force(2).Force);
f = Data{1,1}.Force.Frequency;
Frames = Data{1,1}.Force.NrOfSamples; 

% isolate vertical force
Fges = F1 + F2;
Fvert = Fges(3,:); 

Fvert_links = F1(3,:); 
Fvert_rechts = F2(3,:); 

%% Tapping Test 
 
figure
plot(Fvert_links); 
findpeaks(Fvert_links, "MinPeakHeight",600, "MinPeakDistance",50); 


figure 
plot(Fvert_rechts); 
findpeaks(Fvert_rechts, "MinPeakHeight",600, "MinPeakDistance",50); 

% Number of Taps per Side 
Tap_links = numel(findpeaks(Fvert_links, "MinPeakHeight",600, "MinPeakDistance",50)); 
Tap_rechts = numel(findpeaks(Fvert_rechts, "MinPeakHeight",600, "MinPeakDistance",50)); 

% Number of Taps

Taps = Tap_links + Tap_rechts; 

This is my code so far and the record I use is atteched.

[Mathieu NOE]

hello

maybe I'm wrong, but I felt like you simply need to plot / analyse your last 11250 samples of data

%% Read data
[filename,pathname] = uigetfile('*.mat','Bitte Sprungdaten Mat-Datei auswählen','C:\Users\ifdadmin\Desktop\Local QTM Project\Data\20201007FR02\Running,Squat,CMJ,DJ,Extra\CMJ 2.qtm');
load([pathname, filename]);
% Kraftdaten aus Rohdaten ziehen
Data = struct2cell(load([pathname,filename]));
% Force data 
F1 = (Data{1,1}.Force(1).Force);
F2 = (Data{1,1}.Force(2).Force);
f = Data{1,1}.Force.Frequency;
Frames = Data{1,1}.Force.NrOfSamples; 
% isolate vertical force
Fges = F1 + F2;
Fvert = Fges(3,:); 
Fvert_links = F1(3,:); 
Fvert_rechts = F2(3,:); 

%% take only the last 11250 samples (15 s x 750 Hz)
Fvert_links = Fvert_links(end+1-11250:end); 
Fvert_rechts = Fvert_rechts(end+1-11250:end); 

%% Tapping Test 
 
figure
plot(Fvert_links); 
findpeaks(Fvert_links, "MinPeakHeight",600, "MinPeakDistance",50); 
figure 
plot(Fvert_rechts); 
findpeaks(Fvert_rechts, "MinPeakHeight",600, "MinPeakDistance",50); 
% Number of Taps per Side 
Tap_links = numel(findpeaks(Fvert_links, "MinPeakHeight",600, "MinPeakDistance",50)); 
Tap_rechts = numel(findpeaks(Fvert_rechts, "MinPeakHeight",600, "MinPeakDistance",50)); 
% Number of Taps
Taps = Tap_links + Tap_rechts; 

[Hannah Knoke]

Thank you, that helps a lot! But is there also a possibility to crop the beginning and saying : Start to count the peaks when one is higher than X or an other start event and stop to count after 11250 frames?

[Image Analyst]

Can you post a screenshot of the data plotted?

[Hannah Knoke]

This is the uncutted plot.

[Image Analyst]

What is a "frame"? Is it an element? Can you just do

index1 = find(signal > 600, 1, 'first');
index2 = index1 + 11249; % Index 2 is exactly 15 seconds later.

(Make sure you scroll down to see @Mathieu NOE's answer).

[Hannah Knoke]

Yes a frame is something like an element.

Thank you for your help, I will try this.

Answer 1

[Mathieu NOE]

hello Hannah

I tried to improve my first (very simple suggestion)

here some (hopefully) improvements :

• detection of "valid" peaks based not only on findpeaks (and already used parameters) , but added another step :
• method = all peaks generated by findpeaks are then analyzed in term of time increments (delta), by buffer of 5 samples (with overlap = 1)
• only a full valid (low standard deviation) buffer can trigger a positive ouput , this way we record the first sample index of the valid buffer
• once this first valid peak is defined, we show all peaks from this first value up to the end , 11250 samples down the road.

updated code :

%% Read data
[filename,pathname] = uigetfile('*.mat','Bitte Sprungdaten Mat-Datei auswählen','C:\Users\ifdadmin\Desktop\Local QTM Project\Data\20201007FR02\Running,Squat,CMJ,DJ,Extra\CMJ 2.qtm');
load([pathname, filename]);
% Kraftdaten aus Rohdaten ziehen
Data = struct2cell(load([pathname,filename]));
% Force data 
F1 = (Data{1,1}.Force(1).Force);
F2 = (Data{1,1}.Force(2).Force);
f = Data{1,1}.Force.Frequency;
Frames = Data{1,1}.Force.NrOfSamples; 
% isolate vertical force
Fges = F1 + F2;
Fvert = Fges(3,:); 
Fvert_links = F1(3,:); 
Fvert_rechts = F2(3,:); 

%% Tapping Test 
 MinPeakHt = 600; % MinPeakHeight
 MinPeakDist = 50; % MinPeakDistance
[PKS_links,LOCS_links]=findpeaks(Fvert_links, "MinPeakHeight",MinPeakHt, "MinPeakDistance",MinPeakDist); 
[PKS_rechts,LOCS_rechts]=findpeaks(Fvert_rechts, "MinPeakHeight",MinPeakHt, "MinPeakDistance",MinPeakDist); 

samples_display = 11250; % max samples window size to display (from first valid peak - see below the buffer analysis)
x = 1:length(Fvert_links); % create x axis data

figure
    % remove (eventual) first spurious peaks by looking at buffered data and keeping
    % only values that show a low standard deviation on time increments (delta)
    valid_index_links = buffer_analysis(LOCS_links);
    start_index_links = valid_index_links(1);
    LOCS_links = LOCS_links(start_index_links:end);
    ind_links = LOCS_links(1)-100:min(LOCS_links(1)+samples_display,length(Fvert_links)); % start to display data 100 samples before first peak
    LOCS_links(LOCS_links>max(ind_links)) = []; % remove peaks plot after 11250 samples window
    
plot(x(ind_links),Fvert_links(ind_links)); 
hold on
plot(x(LOCS_links),Fvert_links(LOCS_links),'dr');
hold off

figure
    % remove (eventual) first spurious peaks by looking at buffered data and keeping
    % only values that show a low standard deviation on time increments (delta)
    valid_index_rechts = buffer_analysis(LOCS_rechts);
    start_index_rechts = valid_index_rechts(1);
    LOCS_rechts = LOCS_rechts(start_index_rechts:end);
    ind_rechts = LOCS_rechts(1)-100:min(LOCS_rechts(1)+samples_display,length(Fvert_rechts)); % start to display data 100 samples before first peak
    LOCS_rechts(LOCS_rechts>max(ind_rechts)) = []; % remove peaks plot after 11250 samples window

plot(x(ind_rechts),Fvert_rechts(ind_rechts)); 
hold on
plot(x(LOCS_rechts),Fvert_rechts(LOCS_rechts),'dr');
hold off


% Number of Taps per Side 
Tap_links = numel(LOCS_links); 
Tap_rechts = numel(LOCS_rechts); 

% Number of Taps
Taps = Tap_links + Tap_rechts; 

function out = buffer_analysis(data)
% running (buffered) analysis with one sample overlap : 
buffer = 5;  % nb of samples in one buffer (buffer size) 
stdD_threshold = 50; % time increments (delta) in buffer : valid buffers must have standard deviation below threshold
overlap = buffer-1; % overlap expressed in samples
out = [];
%% main loop %%%%
    m = length(data);
    shift = buffer-overlap;    % nb of samples between 2 contiguous buffers  
    
    for ci=1:(m-buffer)
        start_index = 1+(ci-1)*shift;
        stop_index = start_index+ buffer-1;
        buffer_data = data(start_index:stop_index);  % 
        D = diff(buffer_data);
        if std(D)<stdD_threshold
            out = [out;start_index];
        end

    end
end

[Mathieu NOE]

hello Hannah

hope you're well

have you tried my suggestion ? any feedback ?

all the best

[Hannah Knoke]

Thank you very much!

I will try this and give you feedback if it works for me.

Best

[Hannah Knoke]

It is working, and the result seems right.

Could you explain the function and the loop? Why does the buffer has 5 samples?

[Mathieu NOE]

hello

the idea is that we want to keep only peaks that are sperated by more or less the same time delta

so the function takes the the "raw" peaks from the findpeaks function and will "scan" by buffer of 5 samples if their time distant is constant (minimal standard deviation).

If this is below a given threshold (see : stdD_threshold = 50;) , then we say that these 5 samples are "contiguous" / valid. If not that can be spurious peaks that are still valid from the findpeaks function (according to the parameters we passed) , but may not be "contiguous" with the constrain of being more or less equaly time distant (x axis).

the operation is repeated by blocks (buffers) of 5 samples , with a shift of 1 sample at every for loop iteration, until the last buffer is reached

of course, you can modify the buffer length according to your needs. Increasing the buffer lenght will increase the robustness against small groups of spurious peaks.

choosing the value of 5 samples was my own decision as one of your data was generating 3 initial spurious peaks I wanted to get rid of.

[Hannah Knoke]

Perfect, thank you!

[Mathieu NOE]

as always, my pleasure !