||This thesis addresses a systematic study of monitoring human movements by using a single tri-axial accelerometer in an unsupervised environment. At the beginning, application scenarios and the advantages of using accelerometry are introduced. Fourteen types of human movements, which could be used to analysis the subject’s functional ability in ambulatory monitoring, are then described and set as the target activities which need to be identified. The understanding of TA signals is explored in both theoretical and experimental aspects. Based on this, a data processing schema, which consists of three steps, is proposed. We also discuss the placement issue of using accelerometers and present our own method to place the device. After that, we develop a classification framework for identifying daily human states and activities. Data mining techniques in the field of time series segmentation and query are first introduced to recognizing human movements using accelerometers by our work. Then an unsupervised experiment, which collects data from 8 subjects, is designed to evaluate the framework. The overall accuracy when training and testing data are from different subjects is nearly 90%, which is promising in this area.