Abstract : The aim of this work is to develop a signal processing methodology in order to improve fine studies of the cardiac signal, and specially of P wave, particularly focusing the measurement of shape variations. After a review of cardiac signal characteristics, a description of its physiological and pathological variability and of the different recording techniques, a critical study of cardiac signal processing methods is performed: noise reduction, specific filtering, signal averaging and jitter estimation. An interactive and practical software for ECO, particularly devoted to P wave studies, has been developed, with special algorithms of preprocessing and processing, concerning signal averaging procedure and shape comparison. In fact, the most important part of this work is devoted to a new and personal approach of measuring shape differences of bidimensional signaIs. This method is an extension to two dimensions of the distribution function method (DFM), called DFM-2D. The method is described and it is demonstrated that it can be applied to time-frequency representations of monodimensional signaIs. Through simulations applied on gaussian curves, it is clearly showed that DFM-2D performances are better than DFM-ID ones, on noisy as weIl as non-noisy signals. DFM-2D can also be applied to image processing, as it is proved by examples of histologic sections processing. Precise and early detection of electrophysiological action of drugs (such as Cibenzoline or Quinidine) on ECO waves (P, QRS, T) is an interesting application ofDFM-2D in pharmacology. The presented results demonstrate the abililty of this method to measure shape variations and to distinguish them from wave duration fluctuations.