Forums

Simulation of Laplace Distributed Noise

Started by HardySpicer June 25, 2012
I am doing some testing of blind source estimation schemes and require
higher-order statistical noise.
I only have Gaussian and normal distributions. Is it possible to shape
these in some way to make a Laplace distribution? (to model speech - I
know I could use real speech but I want more control over things).

Thanks


On Monday, June 25, 2012 1:34:01 PM UTC-7, HardySpicer wrote:
> I am doing some testing of blind source estimation schemes and require > higher-order statistical noise. > I only have Gaussian and normal distributions. Is it possible to shape > these in some way to make a Laplace distribution? (to model speech - I > know I could use real speech but I want more control over things). > > Thanks
By finding the inverse CDF of the Laplace distribution you have a mapping from the uniform random variable to your desired random variable. If you want a source I'll dig up one of my books and put it here. I assume you can generate a uniform random variable as well.
On 06/25/2012 10:34 PM, HardySpicer wrote:
> I am doing some testing of blind source estimation schemes and require > higher-order statistical noise. > I only have Gaussian and normal distributions. Is it possible to shape > these in some way to make a Laplace distribution? (to model speech - I > know I could use real speech but I want more control over things).
Well, I would like to suggest: http://en.wikipedia.org/wiki/Laplace_distribution Where everything you need is explained. bye, -- piergiorgio
On Jun 26, 9:14�am, Piergiorgio Sartor
<piergiorgio.sartor.this.should.not.be.u...@nexgo.REMOVETHIS.de>
wrote:
> On 06/25/2012 10:34 PM, HardySpicer wrote: > > > I am doing some testing of blind source estimation schemes and require > > higher-order statistical noise. > > I only have Gaussian and normal distributions. Is it possible to shape > > these in some way to make a Laplace distribution? (to model speech - I > > know I could use real speech but I want more control over things). > > Well, I would like to suggest: > > http://en.wikipedia.org/wiki/Laplace_distribution > > Where everything you need is explained. > > bye, > > -- > > piergiorgio
Indeed it is!! Thanks
On Jun 25, 3:34&#2013266080;pm, HardySpicer <gyansor...@gmail.com> wrote:
> I am doing some testing of blind source estimation schemes and require > higher-order statistical noise. > I only have Gaussian and normal distributions. Is it possible to shape > these in some way to make a Laplace distribution? (to model speech - I > know I could use real speech but I want more control over things). > > Thanks
What is the difference between Gaussian and normal distributions? In any system in which a random number generator simulating a Gaussian random variable is available, surely there is also a uniform random number generator available? Or are you absolutely adamant that a Gaussian random variable _must_ be massaged into a Laplacian random variable; just starting from a uniform random number generator is forbidden? Dilip Sarwate