Data Structure Reference Assignment

Question

Ruud 2021-9-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1449589-data-structure-reference-assignment

评论： Ruud 2021-9-9

When I implement the below it works fine for a single array variable:

Le2=length(GC);
FzGC2=nan(1000,Le2);%consider how to adjust ground contact lengths for different vector lengths during further eval
%Collect last step
for i=1:Le2 %or use N value
    if ~isnan(GC(i,2)) 
    tempFz2 = Fz(GC(i,1):GC(i,2)); %gets ground contact Fz data
    length_tempFz2 = size(tempFz2,1); %find length of ground contact Fz data
    FzGC2(1:length_tempFz2,i)= Fz(GC(i,1):GC(i,2)); %adjusts size of array to store it in
    else
    tempFz2 = Fz(GC(i,1):end); %gets ground contact Fz data
    length_tempFz2 = size(tempFz2,1); %find length of ground contact Fz data
    FzGC2(1:length_tempFz2,i)= Fz(GC(i,1):end); %adjusts size of array to store it in
    end
end

However when I create a struture and try to do the same kind of reference I am getting an 'Index exceeds matrix dimensions' errror on the first bolded line and then the second bolded line if I remove the first while troubleshooting. GC is a 170*2 array with references ranging from 700 to 12000. Both Fz and CoPx are similar sizes - 1 structure with 6 fields (Trials 1-6). Trial sizes within the structure are 30000 or 120000 * 1

for q=1:length(filenames)

Q = size(CoPx.(fnc{q}),1);

%CoPx.(fnc{q})(isnan(fnc{q}))= 0;

for i=1:length(GC)

if ~isnan(GC(i,2))

tempCoPx = CoPx.(fnc{q})(GC(i,1):GC(i,2)); %gets ground contact

length_tempCoPx = size(tempCoPx,1); %find length of ground contact

CoPxGC(1:length_tempCoPx,i)= CoPx.(fnc{q})(GC(i,1):GC(i,2)); %adjusts size of array to store it in

else

tempCoPx = CoPx.(fnc{q})(GC(i,1):end); %gets ground contact Fz data

length_tempCoPx = size(tempCoPx,1); %find length of ground contact Fz data

CoPxGC(1:length_tempCoPx,i)= CoPx.(fnc{q})(GC(i,1):end); %adjusts size of array to store it in

end

CMean = mean(CoPxGC);

A=mean(CMean(1:2:end));%odd number contacts

B=mean(CMean(2:2:end));%even number contacts

%assuming +X is to the right

if A > B

RIGHT=1;

LEFT=2;

else

LEFT=1;

RIGHT=2;

end

%assign [left TD, left TO, right TD, right TO]

end

Can you please assist?

5 个评论
显示 3更早的评论隐藏 3更早的评论

Jan 2021-9-8

编辑：Jan 2021-9-8

在 MATLAB Online 中打开

A hint at first: Simplify your code. The first block can be written as:

Le2   = length(GC);
FzGC2 = nan(1000, Le2);
for i = 1:Le2 %or use N value
    if isnan(GC(i,2))
        cFz = Fz(GC(i,1):end);
    else
        cFz = Fz(GC(i,1):GC(i,2));
    end
    FzGC2(1:numel(cFz, i)) = cFz;  % adjusts size of array to store it in
end

Of course, if you have GC(i,2) and GC(i,2) you can calculate the size of GC(i,2):GC(i,2) easily also:

Le2   = length(GC);
FzGC2 = nan(1000, Le2);
fixGC = GC;
fixGC(isnan(GC(:, 2))) = numel(Fz);  % Care for the NaNs outside
for i = 1:Le2  % or use N value
    a = fixGC(i, 1);
    b = fixGC(i, 2);
    FzGC2(1:b-a+1) = Fz(a:b);  % adjusts size of array to store it in
end

This looks extremly cluttered:

tempCoPx = CoPx.(fnc{q})(GC(i,1):GC(i,2));

Your function contains to many information about the meaning of the data. A more abstract design would simplify the code. E.g. add this on top of the code:

X = CoPx.(fnc{q});

Then use X instead of the compicated expression. This make the code easier to read and it will even run faster.

If the code is cleaner, it gets much easier to find the cause, why CoPx.(fnc{q}) does not have the size you expect. The posted code does not contain the definition of this variable, so the readers cannot find out, where the problem is.

Ruud 2021-9-8

在 MATLAB Online 中打开

Thanks for the response and suggestions around optimization, I'll implement accordingly. the first part of the code was during testing, the second part with the structures is the cricital parts as I need to read in multiple data sets. Definition of CoPx is:

CoPx.(strcat('Trial_',num2str(q)))=c3d.ForcePlate.Data(:,7); %to store in a structure
fnc = fieldnames(CoPx);
%GC:
threshold=1; %set force threshold
C=false; %assume trial starts off ground
n=1;
N=1;
fn = fieldnames(Fz);
L=size(Fz.(fn{q}),1);
GC=nan(1000,2); %ground contacts [TD, TO]
while n<L
    if C %foot is in contact, look for toe-off
        n=find(Fz.(fn{q})(n:L) < threshold,1,'first')+n-1;
        if isempty(n) %cannot find any more
            n=L; %leave while loop
        else
            GC(N,2)=n;
            C=false; %now foot is not in contact
            N=N+1;
        end
    else %look for touchdown
        n=find(Fz.(fn{q})(n:L) > threshold,1,'first')+n-1;
        if isempty(n) %cannot find any more
            n=L; %leave while loop
        else
            GC(N,1)=n;
            C=true; %now foot is in contact
        end
    end
end
end
%Remove NaN from GC
GC(isnan(GC(:,1)),:)=[];

Jan 2021-9-9

在 MATLAB Online 中打开

CoPx.(strcat('Trial_',num2str(q)))

This is hiding the index of an array in the field name. Use arrays with inidices is easier than this indirection.

This looks complicated: Fz.(fn{q})(n:L) < threshold . You can move it out of the loop:

hasFz = (Fz.(fn{q})(n:L) >= threshold);

Now the change point can be found by:

padFz     = [false, hasFz, false];
touchDown = strfind(padFz, [false, true]);
liftOff   = strfind(padFz, [true, false]);

By this way, you do not need a loop.

Ruud 2021-9-9

在 MATLAB Online 中打开

As per suggestions yesterday I've changed all references with .fn{q} to a single letter for easier troubleshooting. I now have:

F = Fz.(fn{q});
X = CoPx.(fnc{q});

CoPxGC now is represented by a 482*170*6 array and CMean 482*1*6 from the code previoulsy shared albeit I am optimzing/simplifying it as we speak. What I realize though is that GC is bascially only holding the last set of values for the iteration so this will have a domino effect on the CoPxGC or FzGC values since it will not have data to reference obvioulsy for the specific trial it is meant to. When I modify GC to GC(N,2,q) it outouts a blank matrix. Forgive me I have minimal experience with dealing with 3D arrays so am learning as I research, modify and test.

Also, I am a bit unclear by the change point suggestion. I substiute to get the below but get a whole different output in CoPxGC and in all scenarios I only get a 2D array for FzGC when adding time normalization.

GC=nan(1000,2); %ground contacts [TD, TO]
while n<L
    if C %foot is in contact, look for toe-off
        n = find(hasFz,1,'first')+n-1;
        if isempty(n) %cannot find any more
            n=L; %leave while loop
        else
            GC(N,2)=n;
            C=false; %now foot is not in contact
            N=N+1;
        end
    else %look for touchdown
        n=find(hasFz,1,'first')+n-1;
        if isempty(n) %cannot find any more
            n=L; %leave while loop
        else
            GC(N,1)=n;
            C=true; %now foot is in contact
        end
    end
end
end

FzGC with time normalization (below returns an index exceeds matrix dimensions when done in this way)

for i = 1:Le %or use N value
    if isnan(GC(i,2))
        cFz = Fz(GC(i,1):end);
    else
        cFz = Fz(GC(i,1):GC(i,2));
    end
    FzGC2(1:numel(cFz, i)) = cFz;  % adjusts size of array to store it in
    FzGC2(:,i,j)=timenorm(cFz,1001)
    
    
    %Time Normalization Function:
function [ output ] = timenorm( data,numsteps)
    %frames=(100/step_size)+1;
    t0=1:length(data);
    t1=linspace(1,length(data),numsteps); %0 to 100% on 0.1% increments i.e calculate what frames are needed if we only have 1001 frames
    output=interp1(t0,data(t0),t1,'spline'); %using frames t0 and Force data interpolate to t1
end

请先登录，再进行评论。

请先登录，再回答此问题。

Data Structure Reference Assignment

5 个评论
显示 3更早的评论隐藏 3更早的评论

回答（0 个）

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

Data Structure Reference Assignment

5 个评论 显示 3更早的评论隐藏 3更早的评论

回答（0 个）

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

5 个评论
显示 3更早的评论隐藏 3更早的评论