Extracting data range from array problem
Show older comments
Hi
I am not very experienced with Matlab and am having a problem trying extract data from an array.
I need to read a csv with 2 coloumns, I then find the peaks in one coloumn, find their index and return the values of the second coloumn at these indexes.
That part I can do.
Then for each of those values returned, (there will probably be between 5 and 10 values) I want to return the 5 values before and after it.
So ideally I would end up with between 5 and 10 arrays of 11 values each of these arrays would need to be accessed indivdually later.
I think I may be able to use cells but not sure how.
When I run the code I have attached it does exactly what I need, BUT it only does it for first value of peakIdx, not all of them. As a result I end up with 1 array of values (t1acdc) but I need a set of values for each value of peakIdx
Any help you be greatly appriciated.
prompt = "How many ts? ";
numTurns = input(prompt)
prompt = "How many seconds sampled before and after? ";
sampTurns = input(prompt)
ipArray=csvread('accel test drive 1.csv');
col1 = ipArray(:, 1);
[peaks,peakIdx] = findpeaks(col1,'MinPeakDistance',300, "MinPeakHeight",10)
acdcPeaks=ipArray(peakIdx,2)
t1acdc=ipArray(peakIdx-sampTurns:peakIdx+sampTurns,2)
2 Comments
Ethan Leonard
on 13 May 2022
It would help to have the csv. Can you add it to the question? Without knowing otherwise, it's possible that it only returns the first one because the others don't have a separation distance greater than 300, in which case it would be working perfectly.
Nigel Davis
on 13 May 2022
Accepted Answer
sampTurns = 5;
ipArray=csvread('accel test drive 1.csv');
col1 = ipArray(:, 1);
[peaks,peakIdx] = findpeaks(col1,'MinPeakDistance',300, "MinPeakHeight",10);
acdcPeaks=ipArray(peakIdx,2);
t1acdc = reshape(ipArray(peakIdx+(-sampTurns:sampTurns),2),[],2*sampTurns+1)
6 Comments
Nigel Davis
on 13 May 2022
Voss
on 13 May 2022
You're welcome!
Let me know if you have any questions about it.
Nigel Davis
on 14 May 2022
The reason that this line
t1acdc=ipArray(peakIdx-sampTurns:peakIdx+sampTurns,2)
only gives you the points around the first peak is because it's using a vector as an argument to the colon operator (peakIdx is a vector, so peakIdx-sampTurns and peakIdx+sampTurns are both vectors). See what happens when you do that:
[21 45 69]:[31 55 79] % example 1
[21 61 91 301]:[31 65 99 302] % example 2
The colon operator only uses the first element of each vector. This makes sense, in a way, because how should the colon operator handle building multiple arrays that are different lengths, like in example 2 there? Maybe it could put them in a cell array and return {21:31 61:65 91:99 301:302}, I suppose, but it does not.
So that's why in my answer I make a matrix of indices
[21; 45; 69]+(-5:5)
and use those as the row index into ipArray with column index 2, and reshape the result.
Now, for the second question, why do you get an error with sampTurns = 80 or higher? It looks like the first peak returned from findpeaks is at index 80, so trying to get elements from the row of ipArray that's 80 rows before that means row 0, which is off the beginning of the array.
sampTurns = 80;
ipArray=csvread('accel test drive 1.csv');
col1 = ipArray(:, 1);
[peaks,peakIdx] = findpeaks(col1,'MinPeakDistance',300, "MinPeakHeight",10)
peakIdx+(-sampTurns:sampTurns)
(Notice the first element in that matrix is 0, which is not a valid index to use.)
With different data or different input parameters in findpeaks, it could happen that the code would try to access a row off the end of ipArray, i.e., a row whose index is greater than size(ipArray,1).
One way to handle these cases and prevent those types of error would be:
- to store those row indices before they're used
- somehow keep track of (or "mark") which row indices are invalid
- replace the invalid values with some valid index
- do the indexing into the rows of ipArray with the "corrected" values
- replace the values in the result t1acdc that correspond to invalid indices with NaN, using the "mark" found before
That would look like this:
% store the row indices:
rowIdx = peakIdx+(-sampTurns:sampTurns);
% create a marker/indicator - a logical matrix saying
% whether each element of rowIdx is invalid:
isBadIdx = rowIdx < 1 | rowIdx > size(ipArray,1);
% replace the invalid row indices with 1
rowIdx(isBadIdx) = 1;
% do the indexing, get the result
t1acdc = reshape(ipArray(rowIdx,2),[],2*sampTurns+1);
% put NaNs in the result where it would've come
% from an invalid row index
t1acdc(isBadIdx) = NaN;
And here's the same thing again, but showing the output of each step. In this case, the only invalid index is the first one (but increasing sampTurns beyond 80 would give you more invalid indices and more corresponding NaNs in the result):
rowIdx = peakIdx+(-sampTurns:sampTurns)
isBadIdx = rowIdx < 1 | rowIdx > size(ipArray,1)
rowIdx(isBadIdx) = 1
t1acdc = reshape(ipArray(rowIdx,2),[],2*sampTurns+1)
t1acdc(isBadIdx) = NaN
Nigel Davis
on 18 May 2022
Voss
on 18 May 2022
You're welcome!
More Answers (0)
Categories
Find more on Matrix Indexing in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
