Ballistocardiogram heart rate detection: Improved methodology based on a three-layer filter

Abstract Voluntary movements and facial expressions adversely affect the accuracy of heart rate detection methods from ballistocardiogram. In this study, the heart rate accuracy was enhanced by improving the selection of region of interest (ROI) and filtering methods. First, the upper edge of the forehead and the tip of the nose were selected as the ROIs. The feature points in the ROIs were selected and their trajectories were tracked. Second, the trajectories were filtered using a limiting filter, a moving average filter, and a Butterworth bandpass filter. Third, a principal component analysis was used to solve the challenge of dimensionality reduction in multidimensional trajectories. Finally, the trajectories were converted to the frequency domain via fast Fourier transform, and the frequency of the maximal spectral amplitude represented the heart rate of the subject. The novel method proposed in this study is more effective than the state-of-the-art methods when subjects exhibit voluntary movement.