Phase space reconstruction for noise-robust respiration rate estimation from PPG signal

Wearable devices that use optical sensing techniques have become increasingly popular in the healthcare industry for detecting diseases at an early stage and monitoring their progression. Multi-parameter monitoring, specifically the extraction of the respiratory signal, is a growing trend due to its simplicity and cost-effectiveness. However, this can result in an inaccurate respiratory rate (RR) due to external noise. To address this, we propose an attractor reconstruction technique with optimized (auto-correlation coefficient) delay 'τ_opt' and a Retract strategy to improve the RR accuracy. The performance is evaluated using the BIDMC dataset, and a mean absolute error of 2.19 and 1.87 breaths per minute are reported for the 32- and 64sec windowed PPG, respectively. The result indicates that the proposed method's RR accuracy surpassed the state-of-the-art techniques. Further, the result infers that τ_opt optimization is insignificant for the PPG signal. Hence, the τ_opt can be selected as 1/3^rd of the signal's average cycle length. In addition, the proposed approach is well within an acceptable range for class II medical devices without losing vital signals.