More Reliable Remote Heart Rate Measurement by Signal Quality Indexes

Abstract

Accuracy of camera-based heart rate $(HR_{cb)}$ measurement is often impaired by artifacts, which leads to erroneous $HR_{cb}$ and reduced confidence in the measurement. To avoid erroneous $HR_{cb}$, we investigated six signal quality indexes (SQIs) from the literature in terms of their effect size and combined them to a novel SQI-filter. All analyses were performed on the “Binghamton-Pitts-burgh-RPI Multimodal Spontaneous Emotion Database” (BP4D+) in three important color channels. Signal-to-noise ratio, average maximum cross correlation of consecutive segments, and relative difference of spectral peaks were the most powerful SQIs. The SQI-filter increased accuracies of all color channels. Largest improvements (up to 60 %) were achieved in the green channel resulting in 80 % accuracy. The overall highest accuracy of 84 % was reached in the hue channel. Motion-rich videos benefited most from the developed SQI-filter. The presented methodology helps to discard distorted signals. This enables more reliable $HR_{cb}$ data in further applications and increases confidence in the measurement.