Design and validation of a respiratory rate estimation algorithm based on photoplethysmography (PPG) signal

Respiratory rate (RR) is an useful vital sign to evaluate a patient’s clinical condition. For this reason, an algorithm based on a linear filter and photoplestimography (PPG) signal is presented. This signal is used by the commercial pulse oximeters and its record uses a non-invasive and low-cost optical technique. The algorithm is characterized by low computational cost, which is suitable to be implemented in devices that have batteries and low computational resources such as wereable devices. Its validation was performed using two public databases available on the web (BIDMC and CCSHS). In total, 541 records with PPG signal were tested. Furthermore, a spectral analisys method was applied to define the filter cuttoff frequency. On the other hand, the root mean square error (RMSE) was used for performance evaluation. Several statistics parameters based on this parameter were computed such as mean (Me), standard desviation (std), median (Med) and interquartile range (IQR). The results obtained in each database, BIDMC (3,07; 1,68; 3,03; 2,13) and CCSHS (3,18; 0,77; 3,10; 1,02), are comparable with other works that use more complex signal processing methods requiring more computational resources. We conclude that the proposed algorithm might be quite relevant for the design of portable monitoring systems, particularly those used in telehealth.