In-Bed Posture Classification and Limb Identification

We propose an algorithm that uses pressure image data to detect a person's sleeping posture and identifies different body limbs. Our algorithm can be used in monitoring bed-bound patients and assessing the risk of pressure ulceration. We used a GMM-based clustering approach for concurrent posture classification and limb identification. Our proposed technique, applied on 9 healthy subjects instructed to sleep in 13 different postures, resulted in 98.4% classification accuracy in distinguishing three main stable sleeping postures. Additionally, 8 limbs in supine and 5 limbs in left/right side postures were identified with the overall accuracy of 91.6%.