CONTEXT

Accurate and reliable heart rate (HR) estimation using photoplethysmographic (PPG)-enabled wearable devices in real-time during daily life activities is challenging.

PROBLEM

A PPG signal recorded using a wearable PPG device is corrupted by motion artifacts. Therefore, the main challenge of monitoring HR in real time is the accurate reconstruction of a clean PPG signal by suppressing motion artifacts.

PROPOSED APPROACH

The proposed algorithm employs the Fourier theory-based Fourier decomposition method (FDM) to suppress motion artifacts and a fast Fourier transform (FFT)-based method to estimate the HR. In this paper, a computationally efficient algorithm that does not require a reference accelerometer signal to suppress motion artifacts to estimate HR in real time during physical activities is proposed.

METHODOLOGY

The noisy PPG signal is decomposed into a desired set of orthogonal Fourier intrinsic band functions (FIBFs). A clean PPG signal is obtained by discarding the FIBFs corrupted with noise and superpositioning the clean FIBFs. Clean FIBFs were further used to estimate the HR.

RESULTS

The proposed method is evaluated by computing the mean absolute error (MAE) and percentage absolute error (PAE) on two publicly available datasets, IEEE SPC (training and test) and BAMI (BAMI-I and BAMI-II). The MAE and PAE values computed with the proposed method using the IEEE SPC dataset were (1.87, 1.71). The MAE and PAE values computed using the proposed method on the BAMI-I and BAMI-II datasets were (1.33, 1.13) and (1.45, 1.17), respectively. The computed MAE and PAE values were more accurate than those of state-of-the-art techniques presented in the literature.

CONCLUSION

Owing to the improved accuracy and speed, the proposed HR estimation algorithm can be implemented in wearable health monitoring devices for continuous and reliable HR estimation in real time. The proposed algorithm can be applied to denoise PPG signals with different sampling rates.