Adaptive comb filters for quasiperiodic physiologic signals

Quasi-periodic physiologic signals such as the electrocardiogram (ECG) and pulse oximetry are sensitive to artifact and noise caused by patient movement, electrical interference, and sensor movement. This reduces the accuracy of algorithms that process these signals such as QRS detectors. The noise cannot be completely eliminated by traditional bandpass filters because the signal bandwidth often has significant overlap with the bandwidth of the artifact. We describe a simple adaptive comb filter (ACF) that attenuates the noise power between the signal harmonics and continuously estimates the heart rate without detecting individual beats. The filter applied to each harmonic has a finite impulse response (FIR), a user-specified bandwidth, and linear-phase to prevent distortion of the signal morphology.