A simple method to determine polymer blend compositions from their Raman spectra is presented. The method is based on expanding linearly the experimentally measured Raman spectrum of the blend, in terms of Raman spectra of pure components. A smooth function has also been included in the linear expansion to take into account the fluorescence interference, inherent to Raman spectroscopy. The coefficients of the linear expansion that give the best fit to the experimentally measured Raman spectrum of the blend are found by using a standard method of parameter estimation (Marquardt-Levenberg). These coefficients are directly related to the blend composition via a simple calibration procedure. Unlike standard methods of processing Raman spectra as deconvolution and curve-fitting-which use Gaussian and/or Lorentzian functions to approximate the spectrum bands-the proposed method does not require either baseline correction or previous knowledge of peak parameters. Also, this method requires less CPU time than deconvolution and curve-fitting procedures, and it is easy to automate. The proposed method has been applied to blends made out of two polymers: Polystyrene (PS) and poly(phenylene oxide) (PPO), to test its precision and consistency. Excellent agreement was found between calculated and expected blend compositions. Also, the reconstructed spectra agree very well with the experimentally measured blend spectra.