Fig. 5

Separation effect based on single ROI for low-quality data: a shows the facial video frames captured under noise condition; b displays the two groups of R/G/B signals based on ROI(I) and ROI(II) and the respective source signals separated by using SOBI algorithm