Mental stress detection using multimodal characterization of PPG signal for personal healthcare applications

Abstract With growing complexities in our society, mental stress has become inevitable in every human life. Long-term mental stress condition could instigate several chronic diseases and thus require its early evaluation. Existing mental stress estimation techniques mostly uses complicated, multi-channel and expert dependent electroencephalogram (EEG) based approaches. Moreover, the respiratory signal presents promising stress-related information, but its acquisition is also complicated and needs multimodal assistance. Hence, in this research a unique approach based on multimodal characterization of the easy-to-acquire Photoplethysmogram (PPG) signal is proposed to assess the stressed condition. Exclusively, the developed algorithm not only uses a primary PPG feature, but also derives the respiratory rate from the same PPG signal via simplified methodologies. The technique is evaluated on the PPG recordings collected from the publicly available DEAP dataset. Efficiency of these easy-to-compute features is then assessed via a simple threshold-based classification technique to categorize the stressed and the relaxed conditions with an average accuracy of 98.43%. Compared to the existing methods, the proposed algorithm not only shows improved performance but the associated simple methodology with minimum acquisition load also justifies its applicability in real-time standalone, personal healthcare applications.