Autocorrelation-Based Algorithm for Heart Rate Estimation Using Finger-Worn Accelerometers

This study presents a power-efficient alternative method for heart rate estimation in smart rings based on ballistocardiography (BCG).

A custom measurement device was developed, consisting of an adjustable ring fitted to the finger and an integrated accelerometer sensor. Additionally, an ECG measurement device was included in the setup to acquire a 3-lead ECG reference signal. An autocorrelation-based algorithm was developed to preprocess the signals and extract heart rate values. Furthermore, several statistical metrics were calculated from the accelerometer signals to evaluate signal quality, with the goal of distinguishing between poor-quality and high-quality signals and to further study the possibility of implementing such a technique in a commercial device.

The proposed algorithm achieved an average mean absolute error of 1.88 bpm (beats per minute) when compared to reference ECG values, with estimations falling within ±5 bpm 93.3% of the time. When the signal quality based filtration was applied, the error decreased to 1.40 bpm, with estimations being within ±5 bpm 96.2% of the time.

The findings indicate that this approach holds significant promise for integration into commercial smart rings, especially in conditions with limited motion artifacts.