Extracting portions of data using indices

Hello,

I have a 1x1 cell titled trim_data as attached below, which contains a variety of gait data. I then also have a list of indexes where heel strike occurs.

I would like to section all the data in trim_data into a new cell titled Gait cycles that contains each portion of data from heel strike to the next heel strike in different structures. If there are 10 indices there should be 9 gait cycles and therefore 9 structures contatining data. I would like the structures to be divided up from the first index value to the second index value and that would be gait cycle 1 and then divided from the second index value to the third index value to make gait cycle 2 and then another structure from index three to index four to make gait cycle 3 etc etc for as many indexes there are.

I have tried using cellfun but I am getting nowhere with it. How would you do this, or where would one even start trying to do this?

I have attached the data below.

If this question is unclear please reach out.

Answer 1

[Mathieu NOE]

hello Alexandra and welcome back !

I tried this , let me know if it's what you needed

the output cell (Gait_cycles) contain now 9 structures with same fields as in the original data , splitted accordingly to your indexes

code :

load('H05_T05_HS_INDEX.mat')
load('trim_data.mat')


S = trim{1}; %  extract struct from cell 
FN = fieldnames(S); % get field names from S

for ci = 1:numel(idx_HS)-1
    T(ci).count = ci; % just to create a new structure T with new field "count" = block data index (for fun)
    start_index = idx_HS(ci);
    stop_index = idx_HS(ci+1);
    
    for ck = 1:numel(FN)
        fn = FN{ck};
        % get data related to that fieldname
        data = S.(fn);
        T(ci).(fn) = data((start_index:stop_index),:); % store data within start / stop indexes
    end
end
    
% store structure T in one cell Gait_cycles
Gait_cycles{1} = T;

[alexandra ligeti]

Hi, Thanks for the welcome back message.

This is exactly what I was after! Thank you once again for your help.

[Mathieu NOE]

as usual, my pleasure !

[alexandra ligeti]

Hi Mathieu,

Was just wondering, if I wanted to do some statistical analysis (RMSE, SE, absolute error etc) for each gait cycle and then store those results in the same structure ''T'' how would I get the data to store in there?

As I would then also like to average all gait cycles to then have another column called average GC which will take all the gait cycles from hs to hs and have a single gait cycle for that person.

How would I go about doing the two sections above? I have attached my attempt below.

close all

clear all

%Load Data

%Load trimmed data with all vicon and ms angles

%Load Hs indexes

load('H05_T05_HS_INDEX.mat')

load('H05_T05_jOINT.mat')

S = trim{1}; % extract struct from cell

FN = fieldnames(S); % get field names from S

for ci = 1:numel(hs_index)-1

T(ci).GaitCycle = ci; % just to create a new structure T with new field "Gait Cycle" = block data index

% %Error titles- did not work when I set them = to ci so just ignoring for now.

% T(ci).err

% T(ci).abserr

% T(ci).mabserr

% T(ci).mmabserr

% T(ci).RMSE

% T(ci).mRMSE

% T(ci).norm_vicon_gc

% T(ci).norm_MS_gc

start_index = hs_index(ci);

stop_index = hs_index(ci+1);

for ck = 1:numel(FN)

fn = FN{ck};

% get data related to that fieldname

data = S.(fn);

T(ci).(fn) = data((start_index:stop_index),:); % store data within start / stop indexes

end

%Finding average of all 10 Vic and MS GC - Neither line works- get dimension error

% average_MS = mean(T(1).Vic(1,:), T(2).Vic(1,:), T(3).Vic(1,:), T(4).Vic(1,:), T(5).Vic(1,:));

% average_Vic = mean(T(ci).Vic);

end

% store structure T in one cell Gait_cycles

Gait_cycles{1} = T;

%Do a loop to plot all ''x'' many gait cycles for MS and Vic- manuall entering values now

%for i = 1:gc

figure()

%Vic

plot(Gait_cycles{1,1}(1).Vic(:,1)); %first GC

hold on

plot(Gait_cycles{1,1}(2).Vic(:,1)); %Second GC

plot(Gait_cycles{1,1}(3).Vic(:,1));

%MS

plot(Gait_cycles{1,1}(1).MS(:,1)); %first GC MS

plot(Gait_cycles{1,1}(2).MS(:,1)); %Second GC MS

plot(Gait_cycles{1,1}(3).MS(:,1));

%Average all 'x num' gc of ms and all vicon

% ave_gc_Vic = mean()

% ave_gc_MS = mean()

%%Error calculations- store in structure how???? At the moment it just calculated the final

%gait cycles error value

for i = 1:ci

abserr = sqrt((T(i).Vic(:,1)-T(i).MS(:,1)).^2); %absolute error of each gait cycle

err = T(i).Vic(:,1)-T(i).MS(:,1);

mabserr = mean(abserr,2);

mmabserr = mean(mabserr); %average error of entire actvity

RMSE = sqrt(mabserr);

mRMSE = mean(RMSE);

norm_vicon_gc = T(i).Vic(:,1) - mean(mean(T(i).Vic(:,1)));

norm_MS_gc = T(i).MS(:,1) - mean(mean(T(i).MS(:,1)));

norm_se = (norm_vicon_gc - norm_MS_gc).^2;

norm_rmse = sqrt(mean(norm_se));

norm_mse = mean(norm_se,2);

norm_mmse = mean(norm_mse);

end

[Mathieu NOE]

hello again Alexandra

I tried a bit to simplify your code regarding the errors (not code errors I mean, but your new data like RMS...)

In order to simplify my task I simply focused on "mean square root" error(s) so that one cycle creates only one value (and not an array , even though we could also store new arrays)

but ok, here is one solution to expand each T structure with two new fields (and their values)

load('H05_T05_HS_INDEX.mat')
load('trim_data.mat')
S = trim{1}; %  extract struct from cell 
FN = fieldnames(S); % get field names from S
for ci = 1:numel(idx_HS)-1
    T(ci).count = ci; % just to create a new structure T with new field "count" = block data index (for fun)
    start_index = idx_HS(ci);
    stop_index = idx_HS(ci+1);
    
    for ck = 1:numel(FN)
        fn = FN{ck};
        % get data related to that fieldname
        data = S.(fn);
        T(ci).(fn) = data((start_index:stop_index),:); % store data within start / stop indexes
    end
end
    
% store structure T in one cell Gait_cycles
Gait_cycles{1} = T;

%%Error calculations- store in structure how???? At the moment it just calculated the final
%gait cycles error value
for k = 1:numel(idx_HS)-1
    
    % error # 1 computation
    err      = T(k).Vic(:,1)-T(k).MS(:,1);
    mRMSE(k)    = sqrt(mean(err.^2)); % average RMS error of one gait cycle
    
    % error # 2 computation
    norm_vicon_gc = T(k).Vic(:,1) - mean(T(k).Vic(:,1),'all'); 
    norm_MS_gc    = T(k).MS(:,1) - mean(T(k).MS(:,1),'all'); 
    norm_mmse(k)  = sqrt(mean(norm_vicon_gc - norm_MS_gc).^2);
    
    % add to existing T structure one gait cycle error info's (mRMSE and
    % norm_mmse)
    T(k).mRMSE = mRMSE(k);
    T(k).norm_mmse = norm_mmse(k);
    
end
 

[Mathieu NOE]

then you can create a single rms value of the 9 errors (9 cycles ) like this (add at the bottom of the code) :

% RMS averaged errors for the entire experiment
mRMSE_entire = sqrt(mean(mRMSE.^2)); % average RMS error for the entire experiment
norm_mmse_entire = sqrt(mean(norm_mmse.^2)); % average RMS error for the entire experiment

[Mathieu NOE]

then either we store these single numbers as news cells

Gait_cycles{2} = mRMSE_entire;

Gait_cycles{3} = norm_mmse_entire;

Gait_cycles = 1×3 cell array

{1×9 struct} {[4.5538]} {[3.9876e-15]}

but I don't really like this option as there is no fieldname like in a structure to tell what's these numbers are

or we add again to the T structure and in order to match with the size of T (1x9) we need to repeat 9 times the same value (is it really a problem ?)

full code :

load('H05_T05_HS_INDEX.mat')
load('trim_data.mat')
S = trim{1}; %  extract struct from cell 
FN = fieldnames(S); % get field names from S
for ci = 1:numel(idx_HS)-1
    T(ci).count = ci; % just to create a new structure T with new field "count" = block data index (for fun)
    start_index = idx_HS(ci);
    stop_index = idx_HS(ci+1);
    
    for ck = 1:numel(FN)
        fn = FN{ck};
        % get data related to that fieldname
        data = S.(fn);
        T(ci).(fn) = data((start_index:stop_index),:); % store data within start / stop indexes
    end
end
    
% store structure T in one cell Gait_cycles
Gait_cycles{1} = T;

%%Error calculations- store in structure how???? At the moment it just calculated the final
%gait cycles error value
for k = 1:numel(idx_HS)-1
    
    % error # 1 computation
    err      = T(k).Vic(:,1)-T(k).MS(:,1);
    mRMSE(k)    = sqrt(mean(err.^2)); % average RMS error of one gait cycle
    
    % error # 2 computation
    norm_vicon_gc = T(k).Vic(:,1) - mean(T(k).Vic(:,1),'all'); 
    norm_MS_gc    = T(k).MS(:,1) - mean(T(k).MS(:,1),'all'); 
    norm_mmse(k)  = sqrt(mean(norm_vicon_gc - norm_MS_gc).^2);
    
    % add to existing T structure one gait cycle error info's (mRMSE and
    % norm_mmse)
    T(k).mRMSE = mRMSE(k);
    T(k).norm_mmse = norm_mmse(k);
    
end
 
% RMS averaged errors for the entire experiment
mRMSE_entire = sqrt(mean(mRMSE.^2)); % average RMS error for the entire experiment
norm_mmse_entire = sqrt(mean(norm_mmse.^2)); % average RMS error for the entire experiment

for k = 1:numel(idx_HS)-1
       % add to existing T structure  the RMS averaged errors for the entire experiment
    T(k).mRMSE_entire = mRMSE_entire;
    T(k).norm_mmse_entire = norm_mmse_entire;
end

[alexandra ligeti]

Ah cheers, thanks for that. Had a few errors but have got it all working now.

[Mathieu NOE]

my pleasure !