A realtime visual feedback system for heart beat signal

Abstract

A commonly used method for heart rate detection is detecting photoplethysmography(PPG) using pulse oximeter, which illuminates human skin and measures minor change in light reflection. As a change of volume of blood flow in vessels can affect the amount of light reflected by the skin, provided that the skin is thin enough and the surrounding noise is minimal. So far attaching camera to finger tips is a popular way to fulfill the conditions stated above. However, more advanced algorithms are developed to achieve "non-contact PPG", which is needed sometimes for more specific conditions in medical use and entertainment. In this project, I am trying to develop an Android application to perform the “non-contact” cardiac pulse detection as well as showing real-time feedback. There are actually more than one possible algorithms, but the general initial approaches are very similar, that is to obtain a light intensity signal from a region of interest (probably human face) over time. In ideal case, the signal should show a sinusoidal waveform with frequency of heart rate. However, that is never the case in non-contact detection since the environment noise is always severe. A large amount of processing is required to extract the hidden heart rate signal form the acquired signal. In this project, I try to make use of filters, Independent Component Analysis (ICA) and fast Fourier transform(FFT) to achieve the goal.