in addition,use mean value curve and min,max curve to sea your observed  phenomena within the pixel and how it correlate

As Hoang said, any satellites can be used to  draw the spectral curve for each pixel. What makes the difference is that the spectral resolution varies from different sensors. That is, the more wavelenth sub-ranges a sensor covers, the more the spectral curve can reflect the spectral feature of ground objects.

Of course, we can, as my predecessors stated. Notice, that the spectral resolution of a curve like this is significantly reduced, unless you use a super- or hyperspectral images instead of multispectral ETM+. It is important to calculate a real reflectance values, basing on the original digital numbers. You should find the formula somewhere in the information attached to your image.

Yes but before that you need to convert digital numbers to radiance and then apply some corrections like atmospheric correction, topographic and so on.

thank you for your answers

i means that i wanna draw  the spectral curve for  single pixel just for each bands not for all the bands, as regards in each band of ETM+ (for example), this sensor acquire spectral sings in rang of 0.45-0.52 micrometer (for band 1), so i think that we can draw the spectral curve for this single band, not for all the bands.  so what is your opinion?

what do you mean by "spectral curve for single pixel just for each bands not for all bands"? if you take a single pixel and a single spectral band, all you get is a one value, so no, you can't draw any curve. Spectral resolution of the sensor limits a resolution of the curve you want to draw, so it is not possible to draw a spectral curve for a whole blue band (1) using just ETM+ image. As I wrote before: you have only 1 value there. You need to use a different image, with higher spectral resolution, providing more than 1 spectral band in a range in question (and more than 2 if you want to drive an actual curve). Try some super- or hyperspectral images, like Hyperion (there you have the list of its bands)

http://eo1.usgs.gov/sensors/hyperioncoverage

Great answers!  Here is another option that may / may not be applicable. Prof Roy (SDSU) has a tutorial that shows how to use WELD data and tools for extracting and manipulating a time series of a given pixel. 

http://globalmonitoring.sdstate.edu/projects/weld/training/time_series.html

thank you so much for your answers

thank you so much for your answers