In AAL context the increasing demand to new developments in healthcare technologies poses inevitability to the thought of easing the patient monitoring in a non-invasive manner, even with commercial and low cost digital cameras as reliable devices that can be employed to measure vital signs. A widely diffused method for contactless heart rate estimation is remote-photoplethysmography from facial video streams. A critical issue in this research area relates to not having always available frontal images of the subject's face. A further critical aspect is represented by the distance from the vision sensor and by highly variable lighting conditions of the environment. In the proposed work, heart rate measurement of the observed subject was estimated through the design and implementation of a software algorithmic pipeline able to manage the aforementioned critical issues. The proposed system is able to detect faces “in the wild” without constraints and consequently improving its usability. The methodology includes data processing algorithmic steps which allow an adequate extraction of features from a ROI that includes pixels identified as facial skin and whose shape and size vary with the distance and orientation of the face. The findings of the preliminary experiments inferred promising results on three cohorts of subjects grouped by age, with a slightly higher average error expressed in terms of Root Mean Square Error for the group that includes elderly subjects, confirming that the approach considered is even suitable for contactless HR measurement of older people.