Recently, a method for the estimation of speed distribution of particles moving in an optically turbid medium has been proposed. The method allows potentially absolute measurement of speed of the particles and can be applied in laser-Doppler perfusion measurements. However, the decomposition technique was limited to short source-detector separations for which the assumption that one photon is Doppler scattered not more than once is fulfilled. In the present paper we show a generalized decomposition technique in which photons can be scattered more than once. We show the theoretical background for decomposition in such a case. We apply a decomposition method for the analysis of laser-Doppler spectra obtained by Monte Carlo simulations. This analysis allows showing noise limits in which the technique can be effectively applied in analysis of measured spectra. We propose an approximated scattering model based on the assumption that for one photon consecutive Doppler scattering events occur on particles moving with the same speed, and we show that this approximation does not influence significantly the uncertainty of the resulting speed distribution. The proposed decomposition procedure is validated in measurements on a physical flow model. The decomposition procedure is also validated by analysis of spectra measured on a physical phantom using laser-Doppler flow meter (Oxford Optronix, UK). A diluted solution of milk was pumped through a tube fixed in an optically turbid material with speed varying from 0 mm s(-1) to 4 mm s(-1). We observed a linear relation between actual speed of milk solution and speed estimated from results of spectra decomposition.