Salat activity recognition using smartphone triaxial accelerometer

Activity recognition using modern smartphones, equipped with sensing capabilities, has got a tremendous attention of research communities in recent years. This paper presents an automated mechanism to recognise Muslims' prayer (Salat) using the tri-axial accelerometer available in current smartphones. Different activities in the prayers are performed in pre-defined pattern, and all the participants move from one state to another at the same time in a congregation prayer. The accelerometer is first used to collect time series data as input, which is then processed to remove redundancy. This clean processed data is then analysed to recognise not only the individual prayer pattern but also the common pattern in group prayer using dynamic time warping. Promising results are demonstrated through experiments where the prayer activity is correctly recognized while other activities such as walking are classified as non-prayer patterns.