Detection and Classification of Postural Transitions in Real-World Conditions

This study proposes a new robust classifier for sit-to-stand (SiSt) and stand-to-sit (StSt) detection in daily activity. The monitoring system consists of a single inertial sensor placed on the trunk. By using dynamic time warping, the trunk acceleration patterns of SiSt and StSi are classified based on their similarity with specific templates. The classification algorithm is validated with actual data obtained in a real-world environment (five healthy subjects and five chronic pain patients); the best accuracy is obtained through using a custom template defined for each subject (>95% for healthy subjects and 89% for chronic pain). Real-world examinations are found to be preferable because after validating results collected in both real-world and laboratory conditions, the controlled conditions' predictions are too optimistic. Finally, the potential of the new method in clinical evaluation is studied in both healthy and frail elderly subjects. Frail elderly participants show a significantly lower rate of postural transitions, longer SiSt duration, and lower SiSt trunk tilt and acceleration compared to healthy elderly subjects. We conclude that the proposed wearable system provides a simple method to detect and characterize postural transitions in healthy, chronic pain, and frail elderly subjects.