Bed posture classification based on artificial neural network using fuzzy c-means and latent semantic analysis

Observation of physical activities and movement patterns is crucial in clinical practice. Current clinical protocols are based on periodic and subjective observations and self-reports. This paper aims to present an efficient monitoring framework for recognizing lying posture and monitoring on-bed activities to assist charting in order to improve patient safety and caregiving efficacy. From pressure images gathered from a developed sensor pad system, an activity scoring mechanism was applied for segmenting rest and movement periods. The fuzzy c-means (FCM) algorithm was used to transform the pressure contours and identify regions of interest (ROI) with high pressure for pressure ulcer prevention. Latent semantic analysis (LSA) extracted the significant features from the transformed ROI images in order to develop an artificial neural network model for posture recognition. Several objective evaluations and a case study were performed to investigate performance. Experimental results show that the average posture recognition rate was 95.89% when the pressure distributions were divided into four clusters. FCM with LSA transformation improved the recognition rate and could be used to locate the corresponding risk regions of bony prominences. The prototype system also revealed encouraging potential in the production of continuous and quantitative information for assisted living charting nursing care.