Open edge–cloud architecture of a wearable stress monitoring device with deterministic biosignal acquisition on FPGA and validation protocol

Wearable devices for recording photoplethysmogram (PPG) and electrodermal activity (EDA/GSR) are promising for objective assessment of stress response, but consumer bracelets often show low reproducibility compared to laboratory standards due to the closed nature of algorithms and the inaccessibility of raw data. The paper proposes an open edge–cloud architecture of a wearable stress monitoring node that combines deterministic multi-channel collection on an FPGA, an ESP32 layer for local pre-processing/communication, and a server-based storage and visualization loop. Metrics are calculated based on 30-second windows (e.g., HR-30s and SCL-30s), and the quality-gate is formalized by window rejection rules, including a threshold of %Bad_PPG ≥ 0.30 (example τ_PPG). The validation protocol uses Pearson and repeated-measures correlation, similar to the reference comparison with Shimmer GSR3+: for Empatica E4, R = 0.596 (HR) and R = 0.681 (SCL), while for Fitbit Sense, R = −0.121 (HR) and R = 0.412 (SCL).