A Preliminary Quantitative Analysis of Contact Force Effects on Peripheral Oxygen Saturation Prediction from Reflectance Photoplethysmography

Reflectance photoplethysmography (PPG) is widely used in wearable devices to estimate peripheral oxygen saturation (SpO2). Although recent studies suggest that contact force (Cf) affects PPG signals, systematic investigations under controlled Cf conditions are not thoroughly conducted. The effect of Cf on PPG-based SpO2detection has not yet been fully investigated. This work presents a quantitative analysis of Cf on PPG under controlled experimental scenarios. A precise force control system was developed to collect the PPG signals under controlled Cf, with a Cf-controlled reflectance PPG dataset, covering Cf from 0.2 to 3.0N. We studied the influence of Cf on PPG signals using PPG-specific metrics. Cf effects were further evaluated through two data-driven tasks. Firstly, we introduced Cf features into an SpO2prediction machine-learning (ML) framework and analyzed its performance improvement. Secondly, we explored the feasibility of using PPG morphological features to classify the Cf levels. The results showed that Cf effectively influenced the PPG signal through motion artifacts and high-frequency noise. Fusing Cf features with PPG features achieved a higher SpO2prediction accuracy of 0.8811% in mean absolute error, compared to both PPG-only ML models and conventional equation-based models. The Cf-level classification results demonstrated that Cf level can be roughly inferred into three classes without using force sensor with AUC of 0.811 on average, providing a potentially low-cost solution for Cf scenario perception. These findings highlight the importance of Cf in reflectance PPG–based SpO2prediction and could potentially support the development of future reflectance PPG systems with Cf corrections.