One of the most important tasks in seismology and applied geophysics is the identification of the different kinds of waves that form a seismic record by means of polarization analysis. In particular, this involves the extraction of body waves (linear polarization) or surface waves (mostly elliptical polarization) from a set of seismic data and which forms a key point in several studies.In this work, a new method of time-frequency polarization analysis based on the stationary wavelet packet transform is developed. The proposed approach identifies and extracts automatically the different waves included in the signal, dependent upon the reciprocal ellipticity. Moreover, the algorithm provides enough information to the user to allow them to also manually select the reciprocal ellipticity intervals, and then extract the corresponding waves of interest contained in the signals.The proposed polarization estimation method and the automatic features extraction algorithm have been evaluated first using synthetic signals, and then applied to real seismic records. Based on the results obtained from both synthetic and real signals, we can conclude that the proposed method correctly identifies and extracts automatically the linearly and ellipticaly polarized waves from the signal, discerning clearly both types of polarization. Moreover, the proposed method is able to identify and extract signals with different kinds of elliptical polarization, allowing us to understand better the characteristics of Rayleigh waves.Index Terms -Polarization analysis, wave identification, seismic signal processing, stationary wavelet packet transform (SWPT).3