Synchronize sensor data from multiple sensors

I have sensor data from four inertial sensors. When collecting data the the .csv files for each sensor have a slightly different amount of rows as some measurements are missed. Now I want to synchronize the four dataframes into one big dataframe that contains the data from all four sensors and is synchronized.

The data looks something like this

Sensor 1 =

SampleTimeFine FreeAcc_X_S1

______________ __________

1 2.0

23 4.0

36 6.0

47 8.0

53 10.0

Sensor 2 =

SampleTimeFine FreeAcc_X_S2

______________ __________

1 3.0

36 5.0

47 6.0

53 2.0

Sensor 3 =

SampleTimeFine FreeAcc_X_S3

______________ __________

1 1.0

23 5.0

47 4.0

53 9.0

Sensor 4 =

SampleTimeFine FreeAcc_X_S4

______________ __________

1 6.0

23 5.0

36 6.0

47 7.0

53 8.0

I want to end up like this:

Sensor 1 =

SampleTimeFine FreeAcc_X_S1

______________ __________

1 2.0

23 4.0

36 6.0

47 8.0

53 10.0

Sensor 2 =

SampleTimeFine FreeAcc_X_S2

______________ __________

1 3.0

23 4.0

36 5.0

47 6.0

53 2.0

Sensor 3 =

SampleTimeFine FreeAcc_X_S3

______________ __________

1 1.0

23 5.0

36 4.5

47 4.0

53 9.0

Sensor 4 =

SampleTimeFine FreeAcc_X_S4

______________ __________

1 6.0

23 5.0

36 6.0

47 7.0

53 8.0

As you can see a new row is added in both table 2 and table 3. The FreeAcc_X is interpolated for these new rows. In the end I want to add these tables into one bigger table. It should look like this:

Sensor_all =

SampleTimeFine FreeAcc_X_S1 FreeAcc_X_S2 FreeAcc_X_S3 FreeAcc_X_S4

______________ __________ __________ __________ __________

1 2.0 3.0 1.0 6.0

23 4.0 4.0 5.0 5.0

36 6.0 5.0 4.5 6.0

47 8.0 6.0 4.0 7.0

53 10.0 2.0 9.0 8.0

What is an easy way to do this?

[Adam Danz]

The solution is similar to the solution I shared in your previous question but we didn't hear back from you on that thread. Instead of using innerjoin you'll use a similar function.

https://www.mathworks.com/matlabcentral/answers/833473-find-numbers-that-are-not-in-multiple-columns#answer_703753

Answer 1

[Star Strider]

If the first columns are not actulal times, as would otherwise be used with datetime or duration, just interpolate.

Considering three of them —

S1 = [ 1                          2.0
      23                          4.0
      36                          6.0
      47                          8.0
      53                         10.0];

S2 = [ 1                           3.0
      36                           5.0
      47                           6.0
      53                           2.0];

S3 = [ 1                           1.0
      23                           5.0
      47                           4.0
      53                           9.0];

Sc = {S1; S2; S3};                                              % Vertically Concatenate Matrices Into Cell Array
Scm = cell2mat(Sc);                                             % Convert To (Nx2) Matrix To Get Unique Times
S_Time = unique(Scm(:,1));                                      % Unique Times

for k = 1:numel(Sc)
    Si(:,k) = interp1(Sc{k}(:,1),Sc{k}(:,2),S_Time);            % Interpolate
end

Snames = compose('Sensor_%02d',1:numel(Sc));                    % Create Variable Names
Combined = [table(S_Time) array2table(Si)];                     % Create Result Table
Combined.Properties.VariableNames = {'SampleTimeFine',Snames{:}}

Combined = 5×4 table
    SampleTimeFine    Sensor_01    Sensor_02    Sensor_03
    ______________    _________    _________    _________

           1              2              3            1  
          23              4         4.2571            5  
          36              6              5       4.4583  
          47              8              6            4  
          53             10              2            9  

This could easily be expanded to the rest of the matrices by including them in ‘Sc’., and differeing Use whatever interpolation method works best. See the interp1 documentation for details.

.

[Dylan den Hartog]

Thank you so much! This is perfect.

[Star Strider]

As always, my pleasure!

[Dylan den Hartog]

My problem changed slightly. Lets say these are the matrices now:

S1 = [ 1 2.0

23 4.0

36 6.0

45 8.2

47 8.0

2 10.0];

S2 = [ 1 3.0

36 5.0

47 6.0

2 2.0

6 3.0]

S3 = [ 1 1.0

23 5.0

47 4.0

2 9.0

4 10.0]

In the end I want to get this:

Combined = 5×4 table

SampleTimeFine Sensor_01 Sensor_02 Sensor_03

______________ _________ _________ _________

1 2 3 1

23 4 number 5

36 6 5 number

45 8.2 number number

47 8 6 4

2 10 2 9

4 number number 10

6 number 3 number

I do not want the SampleTimeFine numbers to be ordered from low to high when using the unique function. So I used the same code but with 'stable' in S_Time = unique(Scm(:,1),'stable');

Sc = {S1; S2; S3}; % Vertically Concatenate Matrices Into Cell Array

Scm = cell2mat(Sc); % Convert To (Nx2) Matrix To Get Unique Times

S_Time = unique(Scm(:,1),'stable'); % Unique Times

for k = 1:numel(Sc)

Si(:,k) = interp1(Sc{k}(:,1),Sc{k}(:,2),S_Time); % Interpolate

end

Snames = compose('Sensor_%02d',1:numel(Sc)); % Create Variable Names

Combined = [table(S_Time) array2table(Si)]; % Create Result Table

Combined.Properties.VariableNames = {'SampleTimeFine',Snames{:}}

And I got this:

Combined =

8×4 table

SampleTimeFine Sensor_01 Sensor_02 Sensor_03

______________ ________________ ________________ ________________

1 2 3 1

23 4 4.13333333333333 5

36 6 5 4.45833333333333

45 8.2 5.81818181818182 4.08333333333333

47 8 6 4

2 10 2 9

6 8.85714285714286 3 9.47368421052632

4 9.42857142857143 2.5 10

It is almost perfect except that the row with SampleTimeFine i = 6 should come after the row with SampleTimeFine i = 4. So like this

SampleTimeFine Sensor_01 Sensor_02 Sensor_03

______________ ________________ ________________ ________________

1 2 3 1

23 4 4.13333333333333 5

36 6 5 4.45833333333333

45 8.2 5.81818181818182 4.08333333333333

47 8 6 4

2 10 2 9

4 number 2.5 10

6 number 3 number

[Star Strider]

In the original problem, the sample times were montonically increasing. This time, they are not.

After doing a bit of experimenting, change the order of the matrices in ‘Sc’ assignment to:

Sc = {S1; S3; S2};

and the order is as you would like it —

S1 = [ 1                          2.0
      23                          4.0
      36                          6.0
      45                          8.2
      47                          8.0
      2                         10.0];
S2 = [ 1                         3.0
      36                           5.0
      47                           6.0
      2                             2.0
      6                             3.0];
S3 = [ 1                         1.0
      23                           5.0
      47                           4.0
      2                             9.0
      4                            10.0];

Sc = {S1; S3; S2};                                              % Vertically Concatenate Matrices Into Cell Array
Scm = cell2mat(Sc);                                             % Convert To (Nx2) Matrix To Get Unique Times
S_Time = unique(Scm(:,1),'stable');                             % Unique Times
for k = 1:numel(Sc)
    Si(:,k) = interp1(Sc{k}(:,1),Sc{k}(:,2),S_Time);            % Interpolate
end
Snames = compose('Sensor_%02d',1:numel(Sc));                    % Create Variable Names
Combined = [table(S_Time) array2table(Si)];                     % Create Result Table
Combined.Properties.VariableNames = {'SampleTimeFine',Snames{:}}

Combined = 8×4 table
    SampleTimeFine    Sensor_01    Sensor_02    Sensor_03
    ______________    _________    _________    _________

           1                2            1            3  
          23                4            5       4.1333  
          36                6       4.4583            5  
          45              8.2       4.0833       5.8182  
          47                8            4            6  
           2               10            9            2  
           4           9.4286           10          2.5  
           6           8.8571       9.4737            3  

That most likely has to do with ‘2’ being common to all of them, while ‘4’ and ‘6’ are not, so specifying the 'stable' sort order means that the order of the matrices themselves in ‘Sc’ is important.

.

Answer 2

[Mohammad Sami]

The easiest way to do this is to import your data as timetables and then use the built in synchronize function. Details and examples are in the documentation.

https://www.mathworks.com/help/matlab/ref/timetable.synchronize.html

[Dylan den Hartog]

The problem is that I want to synchronize the data with the SampleTimeFine which are the sensor measurement numbers. They are not times.

[Mohammad Sami]

If a sample time fine value across different sensor is considered equal, you can use this as a time variable for this purpose.