you can do that using SPtool in matlab 

I did this code 

t=0:.1:20;

n=0:1:1024;

fs=1000;

x=(1/sqrt(2))*sin(2*pi*100*(n/fs))

g=(2/sqrt(2))*sin(2*pi*250*(n/fs))

z=(3/sqrt(2))*sin(2*pi*400*(n/fs))

v=x+g+z

v=awgn(v,20)

sptool

plot(v)

other window will opens while evaluating this code which is sptool 

you can choose the filter you need with the specification of it , also you can see the spectrum of the signals,and signals after the filter

sptool is the solution :)  

You can use Fourier series. After you get the amplitudes and frequencies of all the harmonics (up to a given order), you can selectively change them.

Study digital filters. All causal (filters thatdon't use 'future' values of input or output) will also have a phase associated with them (as well as amplitude filtering). For causal digital filters the best you can have is a linear phase (corresponding to a flat group delay). If the filter is not causal (you can use future samples stored in memory) then you can have a digital filter with phase zero. I lecture DSP and digital filters, but it takes me some 8-9 lectures to cover the whole subject... It does not fit here. :-(

Use FIR or IIR filter commands like (fir1,fir2,impinvar,bilinear,etc.,) to produce filter coefficients for your low pass filter (0 to 15GHz). Then you convolve your input signal and filter coefficients.

Ofcourse I understand that I need to make a convolution of my signal with one filter for amplitude or phase filtering. The real point is how to generate amplitude filter only or phase filter only :) Thanks

symmetries FIR only change amplitude

All Pass IIR filter only change phase

Thanks for these points Ming Zhao

The symmetric FIR filter that Ming Zhao mentioned will have to be non-causal. By symmetric he means one for which the filter coefficients are an even function around the current sample, i.e., a(1) = a(-1), a(2)=a(-2), etc. It is non causal because a(1) will multiply x(n-1) but a(-1) will multiply x(n+1), a future sample, when you write the convolution.

Great points ! Thanks to you, dear Fernando Soares Schlindwein

Try the Wavelet Filtering Techniques. MATLAB has a "Wavelet Toolbox" that you can conveniently use.

Could you send me a tutorial of what you say, please !

If your institution has MATLAB in your lab, the tutorial comes with the license. Check under WAVELET TOOLBOX tutorial.

great I got it !