Pressure Sensor Array Resolution and Sleep Posture Classification

ABSTRACT This study explores methods of classifying sleep posture using pressure sensor data with deep learning models. Considering the importance of sleep and the impact of sleep posture on health status, this study focused on identifying sleep posture. In particular, it sought to improve the system's cost‐effectiveness and ease of use by reducing the number of sensors for posture classification. To achieve this, one‐dimensional array data containing pressure sensor values was transformed into two‐dimensional image data, utilizing ResNet50 and EfficientNet‐B4 models. The experimental results showed that even in the optimized array with a reduced number of sensors, it was possible to maintain over 96% accuracy, proving that optimizing sensors could be a cost‐effective solution while preserving accuracy in sleep posture recognition.