Human Sleeping Posture Recognition Based on Sleeping Pressure Image

Abstract

Pressure ulcer prevention is very necessary for patients with limited movement. Human sleeping posture recognition algorithm is the key technology of intelligent pressure ulcer prevention equipment. In this article, a sleeping pressure image acquisition system is designed. Based on the data collected by the system, we innovatively propose a trunk centerline prediction algorithm based on 1-D Gaussian template. The average prediction deviation of the algorithm is 1.10 pixels (the actual equivalent distance is 1.27 cm). Statistics show that the algorithm has high accuracy for volunteers with different shapes. In addition, we propose a new correction algorithm for sleeping pressure image, which can effectively improve the accuracy of the subsequent sleeping posture recognition algorithm. Finally, the cumulative pressure value of the sleeping pressure images in the vertical direction is extracted as the feature, and a support vector machine (SVM) with sigmoid kernel is used to classify the three basic sleeping postures. Its classification accuracy is up to 97.2%.