Methods and devices are provided for reducing motion artifacts when measuring blood oxygen saturation. A portion of the light having the first wavelength, a portion of light having the second wavelength and a portion of the light having the third wavelength are received. A first signal is produced based on the received portion of light having the first wavelength. Similarly, a second signal is produced based on the received portion of light having the second wavelength, and a third signal is produced based on the received portion of light having the third wavelength. A difference between the second signal and the first signal is determined, wherein the difference signal is first plethysmography signal. Similarly, a difference is determined between the third signal and the first signal to produce a second plethysmography signal. Blood oxygen saturation is then estimated using the first and second plethysmography signals.