Health industry off late has been driven heavily by sensors i.e. accelerometers, magnetometers etc. which has allowed instant medical response to any injurious activity in an indoor/outdoor environment. Among the medical applications of accelerometers, fitness systems have used this component extensively but it still holds prominent room for deployment in an ambient smart home system to monitor daily life. In this paper, a novel accelerometer-based motion recognition system using statistical features have been proposed. Axial components of accelerometer have been processed statistically to produce discriminating features values from each activity. The proposed system was validated against accelerometer dataset and achieved satisfactory accuracy of 79.58% with random forest. The proposed system can be applied to health monitoring systems, interactive games and for examination of behaviors in outdoor and indoor environments.