I am new to Signal processing and want to implement for the measured curve from my experiment. I have a measured curve (like skewed gaussian). I like to change that measured curve to be as close as possible to the target curve (gaussian) (statistically and visually). Visually means, the target curve's width is abit longer than the measured curve when we the two curves are overlapped. Which signal processing should I use? I like to attach a picture, but it is not allowed for the new person. Thanks thomas

  • 1
    $\begingroup$ Perhaps you could post an image elsewhere and provide a link instead. Or some code that can show what you need. It isn't clear from your question. $\endgroup$ – Jason R Mar 23 '12 at 3:39
  • $\begingroup$ Do you want to fit the data to a probability density function? $\endgroup$ – Emre Mar 23 '12 at 4:11

It is unclear what is your application from your question.

If you aim at "gaussianizing" your data -- for example to make better use of statistical modeling, regression, or machine learning methods which work better under a gaussian assumption -- you can use the Box-Cox transform. It consist in applying the following transformation to your data:

$$y = \frac{x^\lambda - 1}{\lambda}$$

$\lambda$ is chosen as to maximize the log-likelihood of a gaussian model given the transformed data - this is a maximization problem usually solved using Levenberg–Marquardt.

I would also strongly suggest posting this to the stats SE.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.