Least-squares support vector machines modelization for time-resolved spectroscopy

This method is based on the time dispersion of light pulses into the scattering medium. The reduced scattering coefficient and the absorption coefficient are usually obtained using numerical optimization technique or Monte Carlo simulation. In this study, we propose to create a prediction model obtained using a semi-parametric modelisation method : the Least-Squares Support Vector Machine. The main advantage of this model is that it uses theorical curve of time dispersion during the calibration step. The prediction can then be performed on different kind of samples such as apples or biological tissue.