eHealth 360°. International Summit on eHealth, Budapest, Hungary, June 14-16, 2016, Revised Selected Papers

Research Article

Towards Stress Detection in Real-Life Scenarios Using Wearable Sensors: Normalization Factor to Reduce Variability in Stress Physiology

Download
303 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-49655-9_34,
        author={Bishal Lamichhane and Ulf Gro\`{a}ekath\o{}fer and Giuseppina Schiavone and Pierluigi Casale},
        title={Towards Stress Detection in Real-Life Scenarios Using Wearable Sensors: Normalization Factor to Reduce Variability in Stress Physiology},
        proceedings={eHealth 360°. International Summit on eHealth, Budapest, Hungary, June 14-16, 2016, Revised Selected Papers},
        proceedings_a={EHEALTH360},
        year={2017},
        month={1},
        keywords={Stress detection Wearable sensors Physiology normalization Machine learning},
        doi={10.1007/978-3-319-49655-9_34}
    }
    
  • Bishal Lamichhane
    Ulf Großekathöfer
    Giuseppina Schiavone
    Pierluigi Casale
    Year: 2017
    Towards Stress Detection in Real-Life Scenarios Using Wearable Sensors: Normalization Factor to Reduce Variability in Stress Physiology
    EHEALTH360
    Springer
    DOI: 10.1007/978-3-319-49655-9_34
Bishal Lamichhane1,*, Ulf Großekathöfer1, Giuseppina Schiavone1, Pierluigi Casale2
  • 1: Holst Center/imec
  • 2: Philips Lighting Research
*Contact email: lamichhane.bishal@gmail.com

Abstract

Wearable physiological sensors offer possibilities for the development of continuous stress detection models. Such models need to address the inter-individual and intra-individual differences in stress physiology. In this paper we propose and evaluate a normalization factor, , to address such differences. is computed using physiological features and the corresponding stress level at a reference point. The proposed normalization factor is evaluated in a dataset obtained from a free-living study with 10 participants, where each participant was monitored for 5 days during their working hours using different physiological sensors. We obtain an average reduction of mean squared error by up to 32% in models with compared to the models without .