Independent component analysis of pulse oximetry signals

This work describes the application of independent component analysis (ICA) to signals from a pulse oximeter, an instrument that measures arterial blood oxygen content. ICA uses estimates of the second- and higher-order joint statistics of the input signals to separate the signal and interference. For time-derivative pulse oximetry signals, the skew of the arterial pulse component is generally much greater than that of the interference, and this can be used to aid the separation. This paper shows an example of ICA used on pulse oximetry signals and then presents simulations to demonstrate that, for pulse oximetry signals, ICA based on third-order statistics of the derivative is indeed superior to fourth-order ICA.