Non-invasive Stress Recognition Framework Using Consumer Internet of Things in Smart Healthcare Applications

Current healthcare models predominantly focus on symptom-based assessments during intermittent physician visits, lacking continuous monitoring capabilities. Acknowledging the uniqueness of individuals and the necessity for ongoing stress monitoring, this chapter introduces a novel wearables-based stress recognition framework tailored for smart healthcare applications leveraging Consumer Internet of Things (CIoT). The framework’s design aims to cater to individual needs and facilitate treatment planning accordingly. The key to the framework’s efficiency is the integration of lightweight machine learning (ML) models at the edge layer, enabling real-time stress prediction while optimizing edge device resources and processing speed. Embracing a consumer-centric approach, this study seamlessly combines edge and cloud layers, with the cloud layer employing a multimodal explainable deep learning model. The performance evaluation against seven state-of-the-art stress recognition approaches demonstrates a 3% accuracy enhancement, highlighting the effectiveness of the proposed methodology. At the edge layer, a Support Vector Machine (SVM) model achieves a remarkable 97% accuracy, showcasing its superiority over other ML models. On the cloud layer, an explainable deep learning model attains 98% accuracy, emphasizing the significance of electrodermal activity and motion signals in stress recognition. By embracing a hybrid edge/cloud approach, the proposed framework delivers non-invasive physiological signal analysis tailored to consumer preferences, surpassing single-layered frameworks in performance and adaptability. This chapter establishes a foundational decision-making criterion for result interpretation and advanced consumer-centric analysis, paving the way for personalized and efficient stress management in smart healthcare ecosystems.