A Pattern-Based Windowed Impulse Rejection Filter for Nonpathological HRV Artifacts Correction

Artifacts in a heart rate variability (HRV) signal can severely distort the extracted time- and frequency-domain parameters, and thus render the information obtained from the signal potentially unusable. In this paper, we propose an algorithm for nonpathological HRV artifact detection called pattern-based windowed impulse rejection (PWIR) filter. This algorithm extends the windowed impulse rejection filter algorithm, we have introduced in our previous publication. Our performance evaluation demonstrates that PWIR compares favorably with respect to the existing algorithms with a true detection rate of 93.80% and a false detection rate of 0.66%. Integral pulse frequency modulation, the most prominent competing algorithm according to our tests, displays a lower true detection rate (87.62%) and a significantly higher false positive detection rate (4.25%). All other algorithms are shown to be significantly less accurate. We also propose an interpolation technique to replace erroneous samples with new ones. The interpolation technique is shown to produce more accurate sample estimates than several existing ones (in terms of maximizing the accuracy of corrected HRV signals when it comes to time and frequency-domain analysis). The combination of the proposed artifact detection and interpolation methods presented in this paper constitutes a nonpathological artifact correction solution for HRV signals.