Sleep comfort analysis using a part-based mixture model with nighttime infrared video

This work investigated a new challenging problem: how to analyze human sleep comfort which is an urgent problem in intelligent home and medical supervision, especially in intelligent temperature control of air conditioners. To overcome this problem, a novel part-based mixture model is proposed to estimate human sleep comfort. Unlike conventional human sleep comfort analysis using uncomfortable and expensive wearable-device, a remote infrared camera and a cheap temperature sensor are used to collect human sleep posture and real-time temperature information. Moreover, a robust sleep posture feature extraction method is firstly proposed to describe sleep comfort not matter human body is covered by a sheet or not. Experiments on a custom-made database demonstrated that the proposed method has promising performance for on-line human sleep comfort analysis.