A wireless networked multi-site triaxial accelerometry system has been devised to estimate activity energy expenditure during daily life. A notable feature of the devised system is the utilisation of activity classification based on multi-site acceleration signals. During performance evaluation tests on a single subject, the proposed method showed higher correlation (adjusted R 2 = 0.952) and smaller estimation deviation (0.76 kcal/min) than two methods without activity classification (multi-site case: adjusted R 2 = 0.924, estimation deviation = 0.892 kcal/min and single-site case: adjusted R 2 = 0.855, estimation deviation =1.18 kcal/min).