Remote Heart Rate Measurement From Near-Infrared Videos Based on Joint Blind Source Separation With Delay-Coordinate Transformation

Noncontact and low-cost heart rate (HR) measurement based on imaging photoplethysmography (iPPG) technology is commonly desired for health care monitoring. However, the usually employed red–green–blue (RGB) cameras are sensitive to illumination variations and cannot work under dark situations. In this study, we propose a novel framework of applying joint blind source separation with delay-coordinate transformation (DCT-JBSS) to evaluate HR from a single-channel near-infrared (NIR) camera in dark situation. First, three facial regions of interest (ROIs) are determined by face detection technique and a single-channel signal is constructed through a frame-by-frame pixel averaging within each ROI. Second, each single-channel signal is transformed into time-delayed multichannel signal through DCT and then treated as a separate ROI signal set. Third, the three ROI signal sets are simultaneously processed by JBSS to derive the underlying shared HR source component vector (SCV), which is usually ordered the first and has the highest correlation across each signal set. Finally, the fast Fourier transform (FFT) is applied to the HR SCV and the corresponding dominant frequency (within the range from 0.7 to 2.5 Hz) with the highest signal-to-noise ratio (SNR) is determined as the target HR frequency. The proposed framework, as well as several other typical iPPG methods, is validated on public DROZY and MR-NIRP databases. The proposed method achieves the best performance, providing a probable way to widen the application of remote and continuous HR measurement during night conditions.