Variability is an inherent feature of human movement, but little research has been done in order to measure such a characteristic using inertial sensors attached to person’s body (wearable sensors). Therefore the aim of this preliminary study is to investigate the assessment of human movement variability for dance activities. We asked thirteen participants to repeatedly dance two salsa steps (simple and complex) for 20 s. We then used a technique from nonlinear dynamics (time-delay embedding) to obtain the reconstructed state space for visual assessment of the variability of dancers. Such reconstructed state space is graphically linked with their level of skillfulness of the participants.