Improved heart rate detection for camera-based photoplethysmography by means of Kalman filtering

Camera-based photoplethysmography (cbPPG) is a recent development in biomedical engineering, which is primarily used to the contactless assessment of subject´s heart rate. cbPPG´s easy acquisition scheme comes at cost of an usually low signal-to-noise ratio. Artefacts such as subject´s movements and light changes impose severe difficulties and have hindered a wider use of the technique up-to-date. This work aims at providing a more reliable heart rate estimation under real application conditions. For this purpose, a Kalman filter complemented by signal quality indexes is proposed. The method is able to reduce the root mean-squared error of estimated heart rates by approximately 37%, while slightly increasing the detection accuracy.