Human movement analysis still suffers from the weakness of the currently used protocols for data collection and reduction. Reliable data comparisons and precise functional assessment require anatomically based definitions of the reference axes and frames, and therefore careful identification and tracking of the landmarks. When impaired children are analysed, the marker-set and other measurement procedures have to be minimised to reduce the time of the experiment and ensure patient collaboration. A new protocol is proposed for the analysis of pelvis and lower limb motion obtained as a compromise between these two requirements. A marker-set is proposed which involves the attachment of 22 skin markers, the calibration by a pointer of 6 anatomical landmarks, and the identification of the hip joint centre by a prediction approach. Anatomical reference frames and joint rotations are defined according to current recommendations. The protocol was assessed by analysing a single child in several repetitions by different examiners, and a population of 10 healthy children, mean age 9.7-years-old. The entire analysis was repeated after subtraction of the offset by static posture angles. The minimum and maximum means of the standard deviations from five examiners of the same child were respectively 2.1 degrees in pelvic obliquity and 6.8 degrees in knee rotation. The minimum and maximum means of the standard deviations from the 10 healthy children were 2.1 degrees in pelvic obliquity and 9.6 degrees in knee internal-external rotation. The protocol is feasible and allows 3D anatomical-based measurements of segment and joint motion and data sharing according to current standards.