The measurement of
human cardiopulmonary parameters based on photoplethysmographic image (PPGI) is
becoming standard due to the increasing need for a contactless measurement
method. There are very few robust, low computational-cost algorithms for mobile
systems with limited processing power. This paper describes a novel method for
heart rate measurement based on PPGI that has a computationally lightweight
algorithm. The proposed method has very low hardware requirements, allowing it
to work online and, furthermore, it has a high tolerance to artifacts. The
approach analyzes the PPGI signal captured by a standard color camera on a
small skin area to estimate heart rate (HR). A pre-filter reduces the offset of
the averaged signal based on Kalman technique. The adopted method uses
short-time Fourier transform (STFT) to find the frequency and phase contents of
the local sections of heartbeats present in the average color intensities of
the Region of Interest (ROI) in the images. The ROI is taken on the person's forehead
region. Tests have been performed in different scenarios to verify the
efficiency of the proposed algorithm. HR was quantified online and compared to
reference measurements based on eighteen persons of different ages. In
addition, the method was used to process data from a large database of videos
with their accompanying electrocardiograms; the results showed a high degree of
agreement between the proposed system and reference measurements. Furthermore,
a more challenging measurement was performed using a test subject walking under
changing illumination conditions. The resulting measurements were more robust
and accurate than those obtained using the method of Independent Component
Analysis (ICA). Finally, plethysmographic signal acquisition is improved by a
robust method of ROI determination and analyzing camera adjustment allows
correct calibration.