Patient overload in hospitals creates a situation where constant supervision cannot be granted to those who need it. In many conditions, early diagnosis of deterioration in the patient's state can save lives. This purpose inspired development of different methods of tracking vital signs such as heart rate, blood pressure, breathing rate etc. These methods have not been integrated into hospital facilities yet.
In this work we test a method to extract heart rate from a thermal video of the patient's face, using an IR camera positioned at the end of the patient's bed.
This method was tested by a simulation based on video of a heat controller and includes three stages. First, the thermal image is translated to temperature units by using linear regression. Then, using the fact that the controller's temperature is constant, the camera and environment noise is estimated using Yule-Walker algorithm for autoregressive models. Finally, the signal from the camera is whitened using the filter from previous stage and correlated with periodical signals of different BPMs (beats per minute). The chosen heart rate is the one with the maximal correlation value.
The simulation results show that heart rate can be estimated accurately when the temperature amplitude is above 0.2 Celsius degrees, and the correlation signals frequency intervals are 1 BPM from 50 to 200 BPM.
These results show that extracting heart rate from thermal video is feasible. This is a foundation for future development of this method with physiological measurements using the same IR sensor.