Pulse oximetry saturation (SpO 2 ) is a main parameter that reflects the condition of the human cardiovascular system. It is best to detect SpO 2 non-destructively with photoplethysmography (PPG). The embedded technique was used to control the dual-wavelength LED illuminating the fingertips, photoelectric sensors to detect reflection data and, then, upload the data to computers, IIR filtering and S-G smoothing filtering on MatLab to process the collected data. The results is shown that the discreteness of the processed data was reduced by 2–6% and the high-frequency noise and motion artifact were removed effectively. It would provide some references for improving the accuracy of the pulse oxygen saturation monitoring system in a complex environment.