Detection of postural transitions using trunk-worn inertial and barometric pressure sensor: application to stroke patients

To better understand how rehabilitation therapy of stroke survivors is transferred in patient’s daily life, activity monitors exist but require multiple wearable devices and may hinder patient’s movements. In this study, the use of a single wearable barometric pressure sensor, placed on the trunk, is investigated as a complementary sensor to inertial sensors for reliably identifying Sit-to-Stand and Stand-to-Sit transitions in daily-life, key components of balance control. The pressure was first converted in altitude and then modeled using a sinusoidal fit. Kinematic features (from the inertial sensor) and altitude features from the model were included after selection in a Logistic Regression-based classifier. Data was collected on 12 stroke patients during a period of 9 hours and involving 345 transitions. A sensitivity of 93.2% and specificity of 89.2% was obtained. The results indicate that the proposed methodology can be used to monitor stroke patients’ lifestyle and evaluate the outcomes of rehabilitation.