Aplicacion de la transformada wavelet al filtrado de senales electrocardiograficas.

摘要

Since it appeared in the 80s, the Wavelet Transform has been used in multiple applications to science and engineering, mainly in the background of processing signals such as biomedical, seismic, audio or radar signals to name but a few.;In this particular case of processing signals, one of the most widely spread application of the Wavelet transform is noise reduction. This precise application is the field on which the present thesis is based.;Thus, the following paper deepens into the process of noise reduction particularly taking electrocardiographic signals, which might have the most utility in the biomedical field.;Firstly, the behaviour of the most common orthogonal Wavelets is studied and characterized against the Gaussian white noise and the real noise in electrocardiographic signals, by using hard and soft as the traditional thresholding functions in methods of noise reduction. Next, a new thresholding function is proposed to improve the performance of the current methods of noise reduction, which are based on the Wavelet transform.;Secondly, to optimize even more the general performance of this tool in long-term signals, methods of automatic signal segmentation based on changes in the noise level are studied. Specifically, a new segmentation method is proposed by using the correlation between changes in the entropy level of signals, estimated by sample entropy, and changes in the power level of noise. This way, thresholds can be calculated locally in areas with stationary noise level, instead of obtaining the global average which leads to a suboptimal thresholding.