One possibility is to use the discrete filter block with denominater = [1] and Numerator are the exponential weights. If you do not want some static gain, then the sum of the weights shoul equal 1, e.g.
Numerator = exp(-[1:10])./sum(exp(-[1:10]));
A hand made solution could exist of a tapped delay block which outputs N delays of the input inside a vector. Then you can multiply your weights to this vector and sum up all elements:
