A Robust Method of Peak Detection in Noisy PPG Signals Using a Structure of IIR Filters

Pulse oximeters are used in hospitals to continuously monitor the blood oxygen levels (SpO2) and heart rates of the patients. Calculation of SpO2 is found by taking the ratio of the AC to the DC components of the photoplethysmograph (PPG) signals measured by the pulse oximeter. Often the AC component is hidden in noise, which conventional filtering methods are unable to effectively extract. This paper describes a robust method of peak and trough detection in the presence of high levels of noise (SNR of 13 dB) using a structure of IIR filters. Two lowpass IIR filters with a cut off frequency difference of 5 Hz are used. The lower frequency (LF) cutoff filter extracts the peaks and troughs of the fundamental pulse frequency and the higher frequency (HF) filter closely tracks the raw data so that when a LF peak/trough is found, the maximum/minimum HF value across the LF pulse period is considered the peak/trough. The cut-off frequencies of the LF and HF filters adjust to the measured heart rate to track the heart rate. Testing and comparison to the typical bandpass filter method produced a specificity of 0.97 and a sensitivity of 0.96, demonstrating the accuracy of the method.