This study is devoted to seismic waves separation which consist of the principal seismic treatment before any physical interpretation. In the first part of this work, the non parametric separation methods (Matrix Spectral, Karhunen-Loève Transformation, Velocity filtering) are studied. The limitations of these methods induced us to develop a parametric method. This approach needs a modelisation which allows us to describe many kind of seismic waves. The parametric wave separation use the Maximum Likelihood Estimator. It is achieved by estimating different parameter (apparent slowness, amplitudes, waveforms, phases and polarisation) which characterise each wave. The different estimation methods and the performances of the parametric separation are studied in the second part. The results of this study are applied, in the last part, to the analysis of weil survey experimentation as weil as surface waves.
