This paper proposes a circadian life pattern model and anomaly detection method for the elderly living alone by using a remote monitoring system. The proposed circadian model is composed of a set of fuzzy sets from time information of features such as staying of each room, going-out event, sleep behavior. The feature can be considered as components of daily living pattern of a resident. The fuzzy sets represent a statistical repetition of the components. Based on the derived life pattern model we propose a measure called the degree of anomaly that can indicate behavioral deviations from usual habits. In order to verify the proposed method, we installed the remote monitoring system into three houses and collected data during over three months, then analyzed the performance of the proposed method.