Pulse oximetry through spectral analysis

Signal processing algorithms for pulse oximetry waveforms are typically completed in the time domain utilizing analog filtering and moving average techniques. This investigation evaluated the use of spectral analysis as an alternative algorithm for the signal processing of pulse oximetry waveforms. Fast Fourier transformation of eight separate red and infrared waveform data segments was accomplished via a custom C-program. Magnitudes of spectral peaks in the cardiac spectrum were utilized to compute an SpO2 value. In this limited trial, spectral analysis derived SpO2 values were at least as accurate as time series SpO2 values. However, further investigation is needed to further define the limits of spectral analysis on pulse oximetry waveforms with unusual respiratory variations, electrical noise, and motion artifact.