Monitoring the habits of elderly people is a great challenge in order to improve ageing at home. Studying the deviances from or the evolution of regular behaviors may help to detect emerging pathologies. Regular patterns are searched in the data coming from sensors disseminated in the elderly's home. An efficient algorithm, xED, is proposed to mine such patterns. It emphasizes the description of the variability in the times when habits usually occur, and is robust to parasite events. Experiment on real-life data shows the interest of xED.