A Novel Body Posture Recognition System on Bed

Diseases such as positional obstructive sleep apnea and pressure sores are closely related to body postures on a bed. To estimate body postures reliably and comfortably, we proposed a novel classification system using unconstrained measurements of ballistocardiogram (BCG) signals. A flexiblepiezo-electric polymer film sensor was embedded in a mattress to collect BCG data. The amplitude features based on the morphology of BCG were applied to Bayesian classifier with piecewisesmoothing correction. Twelve healthy subjects participated in the experiment. The final average prediction accuracies of four common body postures (supine, left lateral, prone and right lateral) on the bed all exceeded 97%, and a list of kappa coefficients calculated from classification results of subjects demonstrated almost perfect agreement. Overall, the developed system represents a brand new thinking in building an unobtrusive solution for body posture monitoring in our daily lives.