Abstract: Traditional activity recognition methods based on acceleration sensors generally have the assumption that the orientation and placement of sensing devices are fixed．But the recognition performance will be greatly affected when this assumption fails．However,mobile phones,the most widely used sensing devices in pervasive computing environments,are usually placed at unfixed orientation and placement．In this paper,an activity recognition method based on indepen-dent acceleration sensor orientation and placement was proposed to resolve this problem．First,the original 3D acceleration signals are processed into one-dimensional signals．Then,the concept ‘Motif’ from bioinformatics is borrowed to extract position-independent patterns from one-dimensional signals．Finally,Vector Space Model (VSM) based on extracted patterns is built to conduct activity recognition．Experimental results show that recognition rate of the method reaches to 81.41% under the condition of unfixed orientation and placement of sensing devices.

Key words: Activity recognition,Sensor placement,Accelerometer,Motif discovery,Pervasive computing