We propose the use of a retrieval procedure for time-resolved near-infrared tissue spectroscopy based on the 'optimal estimation' method. The aim of this retrieval method is to obtain an improved estimate of the target parameters compared with standard nonlinear least-squares routines, since the inverse problem dedicated to retrieve the optical properties of tissue is ill posed. A priori information on target and forward model parameters is used, so that a larger number of target parameters can be retrieved, and/or a better accuracy and precision can be achieved on the retrieved target parameters. The procedure has been tested on time-resolved simulated experiments generated, using solutions of the diffusion equation and with solutions of the radiative transfer equation reconstructed with Monte Carlo simulations. The results obtained show that, by using a priori information on target parameters, we have a smaller difference between retrieved values and true values, and lower retrieved error bars. Similarly, a more correct estimate of the errors of the forward model parameters improves the retrieval of the target parameters.