A Pulse-Width Modulation Model of Respiratory Signatures Obtained Through Facial Thermograms

Thermal imaging has found prominence in estimating the respiration rate (RR) in a non-contact fashion. The estimation is based on the thermal signature of the nostril regions, which varies with inhalations and exhalations. In our previous research, we devised an algorithm to estimate the RR from the thermal signature across frames. In this paper, we have two contributions. The first is to model the respiratory signal (RS) in a pulse-width modulating (PWM) square wave. We find the model parameters from the experimental data and validate our model with facts from the literature. Next, we characterize the noise obtained by subtracting the respiratory signature from the PWM model. We have used the probability density function (pdf) and power spectral density (psd) for the characterization. We experiment with 30 healthy students. This research would provide insight into the efficacy of using thermal imaging techniques for RR estimation.