Summary:
This thesis deals with the problem of measuring the heart rate of humans. Initially, we present the physiology and the functioning of the human heart and also the most common methods to measure the heart rate. These methods typically require the use of slightly or even particularly invasive devices. Νext, we survey research efforts that deal with the problem of non-invasive heart rate measurement in humans, either by using a typical personal computer, equipped with a web camera, or a smart mobile phone. Through this survey, we end up to the general principles of the methodologies that are followed in both categories. Then, we choose to implement a methodology that measures the heart rate using visual features that are extracted and tracked from the face. Each feature's movement in the y-axis corresponds to a time series. Upon temporal filtering and principal component analysis, we choose the most characteristic series. Finally, we heuristically select the one that corresponds to the heart rate, which is computed upon analysis in the frequency domain. This methodology is experimentally evaluated in terms of both accuracy and usability and is compared to a "smart" bracelet and a mobile application that uses the camera and the flash of a smart mobile phone.
