Skip to content

Signal-quality indicators

Every indicator returns a raw value, normalized quality in [0, 1], an artifact family, and diagnostic evidence. Thresholds are profile-specific engineering priors and are superseded by fitted fusion weights.

Spectral power ratio

Compares power in the profile’s expected physiological band with total power. Welch’s estimator is used for sufficiently long windows. This is related to the pSQI family described by Clifford et al. (2012).

Spectral entropy

Computes normalized Shannon entropy of the power spectral density. Extremely concentrated or diffuse spectra can both be implausible; profile thresholds map the raw value to quality without hiding it.

Mains interference

Measures power near 50 or 60 Hz and harmonics relative to neighboring or total power. The analysis tests the configured mains frequency and reports the evidence used.

Clipping, flatline, and amplitude

Detects repeated samples near observed rails, long runs with negligible first-difference, and implausible RMS. Because recording units differ by hardware, amplitude evidence is treated cautiously unless the profile supplies meaningful bounds.

Baseline wander

Measures low-frequency power beneath the profile cutoff relative to total power after removing the mean. It is evidence of drift or motion, not proof of the physical source.

Distribution shape

Skewness (sSQI) and kurtosis (kSQI) summarize waveform shape. These are weak indicators individually and receive modest fusion weight.

Periodicity and beat agreement

Autocorrelation peak prominence estimates quasi-periodic consistency. For ECG and PPG, two independent peak detectors are compared using tolerance-matched F1 agreement (bSQI), and aligned beat shapes are compared with their robust template.

References

  • Clifford, G. D., Lopez, D., Li, Q., & Rezek, I. (2012). Signal quality indices and data fusion for determining clinical acceptability of electrocardiograms. Physiological Measurement, 33(9), 1419–1433.
  • Li, Q., Mark, R. G., & Clifford, G. D. (2008). Robust heart rate estimation from multiple asynchronous noisy sources using signal quality indices and a Kalman filter. Physiological Measurement, 29(1), 15–32.
  • Moody, G. B., Muldrow, W. E., & Mark, R. G. (1984). A noise stress test for arrhythmia detectors. Computers in Cardiology, 11, 381–384.
  • Silva, I., Moody, G. B., & Celi, L. (2011). Improving the quality of ECGs collected using mobile phones: the PhysioNet/Computing in Cardiology Challenge 2011. Computing in Cardiology, 38, 273–276.