This paper applies digital PLL (Phase-locked loop) approach to the body movement classification problem. PLLs have been used efficiently in telecommunication to retrieve modulated signals from the background noise. Acceleration sensor signal processing often require the same kind of noise tolerance and the relevant movement patterns are frequently periodic, e.g. walking. The paper presents the PLL-based algorithm developed for step counting on a mobile phone and evaluates it against a commercially available wearable step counter.