AI 기반의 향상된 수면자세 분류

Abstract

Frequent tossing and turning reduces sleep quality. One way to measure tossing and turning during sleep is to monitor changes in sleeping position. There are cases where changes in sleeping posture are recognized by deep learning using data collected from pressure sensors placed on the mat, the more pressure sensors, the better the accuracy of the model. However, as the number of pressure sensors increases, the cost of data collection increases. To solve this problem, this paper proposes a deep learn ing model that can classify sleeping posture using fewer pressure sensors compared to previous studies. In this model, 13 pressure sensors are placed on the human shoulder, the top of the mat is divided into five zones according to the position of the pressure sensors, and the data is preprocessed in five sepa rate zones. This means that it is possible to classify sleeping posture using fewer pressure sensors than previous studies. To evaluate the performance of the proposed model, we conducted an experiment to compare the model's classification with the actual posture based on the collected data. As a result, the proposed model correctly evaluated 100% of the 176 data. Therefore, the performance of the proposed model was proved.