Noncontact Heart Rate And Respiration Rate Measurements Using A FastICA Algorithm

Recent years, many cities in our country, especially Beijing, are suffering from hazy weather. Air pollution has done great harm to human body’s respiratory system, with a yearly increase in respiratory diseases. Meanwhile, the haze makes a high incidence of cardiovascular diseases. As an important characteristic index in the cardiovascular disease’s examination, the heart rate(HR) is of great importance to human health. Hence fast and convenient heart rate measurement becomes popular in our daily life.A fast independent component analysis algorithm(FastICA) is introduced to image processing based on the video recordings of human face in this dissertation. In the normal environmental circumstances, we record a digital color video of human face, from which human faces are detected. Then, we extract the heart rate signal from the face images and compute the period of the signal, realizing a simple non-contact heart rate measurement. The method is significantly convenient for developing personal health care and reducing unknown risks. Without the need of touching human skin, the heart rate can be measured indirectly with our method. Compared with traditional contact heart rate measuring methods, our noncontact heart rate measurement method has many advantages such as non-invasiveness, simplicity and high efficiency. Based on PhotoPlethysmography(PPG), a theoretical background for noncontact detection of heart rate using face videos is reviewed first. Then the fast independent component analysis(FastICA) algorithm is applied to blind separation of the observation signals. The algorithm can decompose a group of original observation signals into one group of independent signals by optimally and linearly decomposing the mixture signals. Subsequently, many signal processing methods like correlation analysis, smoothing, filtering, and interpolation are used for automated computation of the HR. Due to the fact that the human face images also has information about another human physiological parameter — the respiratory rate(RR), by analyzing the periodic HR signal and the power spectral density estimation based on the Lomb periodogram, the RR can be estimated from the power spectrum. Experimental simulations show that the heart rate measurement method in this dissertation has a higher reliability compared with the measurement data from the Fingertip Pulse Oximeter.