Spectral analysis through filter banks aplied to preprocessing oriented to thresholding of pulse oximetry signal

Abstract

The pulse oximetry signal (SPO2), allows the calculation of the oxygen saturation level in the blood, and it is acquired over the index finger of the patient. Under normal conditions, the variations in SPO2 have correlated with heart rhythm and its maximum value is in phase with the R wave in electrocardiographic signal (ECG). This property enables the SPO2 signal to be the basis for an alternative method for estimating the instantaneous heart rate. For measuring the instantaneous heart rate from the SPO2, it is necessary to carry out a signal thresholding process for detecting peak values in phase with the R-wave of the cardiac complex. In this paper, an iterative solution method is proposed to establish the cutoff frequency selection for the design of digital filters that allow detection of the maximum values of the signal pulse oximetry. The results obtained from the implementation of filter banks, demonstrated their ability to obtain versions of the pulse oximetry signal and frequency values of the spectral components, associated with the maximum values of the SPO2. Experiments used the CAPNOBASE processed signals database, which contains SPO2 and ECG signals, acquired simultaneously. The results allowed to verify that the filter bank allows to select the appropriate version of SPO2 signal with positive peaks, in phase with the R wave of the ECG signal.

Keywords

• Pulse oximetry
• filter bank
• spectral analysis
• heart rate