Comparación de Técnicas de Procesamiento de Video en Diferentes Espacios de Color para la Estimación de la Frecuencia Cardíaca

Abstract

This study aims to compare different analysis techniques to determine which of these provides better results in estimating heart rate through video processing. The techniques, in general, consisted in the extraction of signals of color intensity variation throughout the video on a region of interest (ROI) of the face, to then determine the dominant frequency of these signals within a possible range of frequencies. Three ROIs were evaluated to find out which region provides more information on cardiac activity: the forehead, the zygomatic region and the area of the face that ranges from the lower part of the eyes to the chin (polygon area). On the other hand, this project also seeks to determine which of the following color spaces allows to obtain better results: RGB, YUV, HSI, HSV and CMY. The best result was obtained when applied the Fourier Transform to the signal extracted from the green channel (of the RGB color space) of the video images, with a root mean square error of 1.5 beats per minutes and a Pearson correlation coefficient of 0.993.