reconstruct DWT with detail 2 to detail 5 for 1-D signal

I'm using db8 DWT for my 1-D signal How can I keep details between detail(2) to detail(5) and remove the others and reconstruct the signal with those keeped?

Answer 1

[Wayne King]

[C,L] = wavedec(x,8,'db8');
Cnew = zeros(size(C));
indices = cumsum(L);
Cnew(1:L(1)) = C(1:L(1));
Cnew(indices(4)+1:indices(9)) = C(indices(4)+1:indices(9));
xnew = waverec(Cnew,L,'db8');

Answer 2

[Wayne King]

Hi I'm assuming you want to keep the level 5 scaling (approximation) coefficients.

Let x be your signal. You can zero out the finest scale details and reconstruct your approximation with the following code.

    [C,L] = wavedec(x,5,'db8');
    C(L(end-1)+1:end) = 0;
    xnew = waverec(C,L,'db8');

Are you sure what to just zero out all the finest details? Consider using wden, or at least wthresh on the finest scale details.

[claire]

Actually I'm decomposing my signal in 8 levels using this command :

[swa,swd] = swt(X,8,'db8'); which decomposes my signal in 8 approximations and details

Is there a way to reconstruct DWT by keeping details from level 2 to level 5 and removing others?

[Wayne King]

If you are only keeping details 2 to 5, why do you decompose your signal down to level 8? what is the point of that?

[claire]

I need to keep later details from level 5 to level 8 to extract others features that I need from my signal because small pics in the signal are better detected in high levels