In this work, we demonstrate a SpO2 monitoring technique based on flexible–window filtering, to get rid of various interferences on the PPG signals, such as ambient light scattering, reflections, and motion artifacts. The filtering window is self-adapted to the characters of the input PPG waveform, rejecting various frequency components of interferences. The proposed technique is validated by in-vivo testing on the human body through a prototype, which is also compared to the conventional fixed-window methods. Measured results show that the proposed technique achieves an average MAPE which is 46% higher than the conventional method, indicating a much stronger immunity to various interferences.