Predictive Coding with Simultaneous Extraction of Pulse and Respiration Rates from PPG Signal for Energy Constrained Wearable Devices

Abstract

Nowadays, wearable sensors or portable devices have great potentials for real-time monitoring of health and fitness of an individual but they are constrained with limited battery power. Therefore, exploring lightweight signal processing technique is highly demanded for accurately measuring the pulse rate (PR) and respiration rate (RR) from the photoplethysmo-gram (PPG) signal in addition to the data compression in order to reduce or even eliminate the need for frequent charging of devices and replacement of batteries. In this paper, we present a lightweight unified predictive coding framework for achieving simultaneous data compression, PR and RR extraction from the PPG signal. Evaluation results demonstrate that the proposed unified framework can achieve compression ratio of 4:1 with energy saving of 52.38 %. For PR estimation, the method had mean absolute error (MAE) of 1.20 (bpm), Pearson coefficient of 0.9829 and Bland Altman ratio of 5.37. The RR estimation had promising MAE results of 3.1 (1.5-5.6 for 25th-75th percentiles) and outperforms the existing methods.