Heart Rate Estimation Using the EVM Method, the FFT and MUSIC Algorithms Under Different Conditions

Abstract

The detection of vital parameters is one of the first and foremost task to be performed on a patient to evaluate her/his health status. Among the vital signs, a great importance is assumed by the heart rate, which is defined as the number of heartbeats measured in a minute. Heart rate is generally measured using an electrocardiograph, which allows an accurate recording of the cardiac activity, but requires the placement of some electrodes in specific points of the body. Such an approach involves a direct contact between the device and the patient’s skin, which is often uncomfortable and impractical. In this paper we propose a contactless approach based on RGB video for heart rate extraction, avoiding any type of direct interaction. For this purpose, we exploit videos of the subjects’ faces recorded via smartphone and we then apply the EVM method to estimate the heart rate obtained in a comfortable and non-invasive way. The results obtained with two different methodologies, based on Fast Fourier Transform (FFT) and on the MUltiple SIgnal Classification (MUSIC) algorithm, have been compared to three devices with clinical validity, i.e., a pulse oximeter, a Polar heart rate strap and a blood pressure monitor, which guarantee accuracy at a reduced cost and are easily commercially available. The high accuracy of the developed system is proved by the small difference achieved between the values measured with the developed contactless technique and those obtained with the wearable devices, which results in an error between 1.07$\%$ and 4.67$\%$, regardless of the ambient light conditions under which the videos were captured.