Computer vision-based navigation for virtual reality applications is always at risk of a tracking failure. An extension for a navigation system is proposed, which recognizes agents motion and stillness states from accelerometer data.