HEART RATE ESTIMATION USING HERMITE TRANSFORM VIDEO MAGNIFICATION AND DEEP LEARNING

Abstract

The primary objective of this project, titled Heart Rate Estimation using Hermite Transform Video 
Magnification and Deep Learning, is to develop an efficient, non- contact method for accurately estimating 
heart rate from video recordings. Traditional heart rate monitoring techniques, such as electrocardiograms 
(ECG) and photoplethysmography (PPG), require direct physical contact with the individual. This project aims 
to address the limitations of contact-based methods by proposing a video- based approach that leverages 
Hermite Transform, video magnification, and deep learning to capture subtle physiological signals indicative 
of heart rate. 
In this study, the Hermite Transform is utilized as a feature extraction tool to detect and enhance fine 
details within video frames, isolating subtle changes that correspond to blood flow and cardiovascular pulse. 
By combining the Hermite Transform with video magnification techniques—particularly Eulerian Video 
Magnification (EVM)—we amplify the faint colour and movement changes in the skin that are linked to 
heartbeats. These magnified signals are then fed into a deep learning model designed to identify and quantify 
the periodic changes associated with heart rate. This hybrid approach allows for high-sensitivity detection of 
the pulse signal without direct contact. 
Extensive experiments have been conducted on public datasets to validate the proposed method, 
comparing its performance with traditional heart rate estimation models. The results demonstrate that this 
method achieves high accuracy and reliability, particularly under varying lighting conditions and with minimal 
preprocessing. Moreover, this approach presents a scalable solution for applications in telemedicine, remote 
health monitoring, and wearable technology. The findings of this project could contribute to advancing non
invasive health monitoring, opening doors to new possibilities in real-time, accessible patient care solutions.