How to Evaluate Signal Quality of ear-ECG?

Hearable devices enable heart rate (HR) estimation through electrocardiogram (ECG), measured from the ears. However, ear-based ECG poses challenges due to its lower amplitude than standard ECG and susceptibility to interference from EEG and movement, which makes its analysis inaccurate. To address these challenges, we undertake a comprehensive evaluation of ear-ECG signal quality using 13 metrics adapted from standard ECG. Theses are based on: higher-order signal statistics, frequency-based indexes, template-based indexes. In addition performance of the deep matched filter detector -R-peak detector specifically dedicated to analysis ear-ECG -is used as a metric. In this study, we utilised data from 40 healthy subjects under the age of 35, recorded while sitting for 5 minutes. Based on the sensitivity of R-peaks detected from ear-ECG referenced to standard ECG, we categorize the signal segments into three classes. Building on this, we propose a simplified approach using four signal quality metrics: 1) the output of a deep-matched filter for R-peak detection in ear-ECG, 2) the ratio of power in the QRS frequency band to the ECG signal band, 3) the ratio of ECG signal band power to the total power, and 4) the ratio of baseline drift power to ECG signal band power. This a practical method for assessing ear-ECG signal quality using deep-matched filters achieves both a lower error rate, and offers a balanced solution which combines low computational complexity, interpretability, and robust performance.