Let's say you have a sinusoid x_ref. And you have done also some processing on some other signals to produce an estimate of x_ref, called x_i. What would be a good way to find the error between the two signals?
I can do a point by point mean absolute error, but is there a good way to normalize this so that I could represent the error as a percentage or something that anyone would understand?
I can tell you that mean squared error is used more widely than absolute error in signal processing community.
Besides that, the first thing which comes to my mind is that, first you can compute the mean squared sum of the sinusoid, second you find the mean squared error sum of the two signal and finally you compute the ratio of them.