Signal Quality Indices based on Skewness of Peak-Peak Interval for Wearable ECG Devices

Wearable devices with continuous ECG recording have become more popular because of their comfort and compactness. However, its measured ECG signals are less reliable due to the effect of muscle noise, baseline noise, and motion artifacts. If all the signals are sent, including unusable data, it drains the device's battery which has a limited capacity. On the server side, the costs, such as power consumption, memory usage, and expense can also be wasted, especially in the trend of all data being uploaded and processed on cloud-based applications. There is a demand to classify the quality of the signal in order only to send good signals to the server. This paper proposes a novel Signal Quality Indices (SQI) based on the skewness of peak-peak interval. Peaks are defined by a value two times higher than an adaptive threshold, which is calculated using the Exponentially Weighted Mean-Variance (EWMV). The results show that the proposed algorithm has higher sensitivity than previous SQI methods, such as template matching and high-order statistical methods based on the ECG amplitude for both noise-free and noisy signals. The proposed method has a sensitivity of 99.02% for noise-free signals, 97.69% for noisy signals, and 96.37% for a database from the Physionet challenge, which contains data collected from mobile phones. The results prove the high feasibility of the algorithm applied to ECG wearable devices.