CASINOR: Combination of adaptive filters using single noise reference signal for heart rate estimation from PPG signals

Abstract

The accuracy of the heart rate (HR) estimated from the photoplethysmography (PPG) signal is always skeptical due to the ubiquity of the motion artifact (MA) in the signal. To ensure the HR estimated from the PPG signal is accurate, a new algorithm is proposed for the removal of MA component from the PPG signal using accelerometer signals. In this paper, the combination of adaptive filters using single noise reference signal referred to as CASINOR is proposed. The recursive least square (RLS) and normalized least mean square (NLMS) adaptive filters are used for denoising the PPG signals. The selection of single noise reference signal is based on the power value of the accelerometer signals. The MA-reduced PPG signals obtained from these RLS and NLMS filters are combined to a single PPG signal using sigmoid function, and the HR is estimated. The estimated HR is refined by using phase vocoder. The proposed method resulted in less HR estimation error of 1.92 beat per minute (BPM) on 23 PPG datasets (IEEE SP CUP 2015). Reduced error value, less computational time and improved accuracy augment this technique to be implemented in wearable devices.