The invention relates to a posture data set collection processing method used for wearable motion sensing game equipment. The method comprises the following steps that 1, posture expression is defined; 2, discrete target postures are generated at random according to the posture expression definition; 3, data samples of the discrete target postures are collected; 4, continuous target postures are generated at random according to the posture expression definition; 5, data samples of the continuous target postures are collected; 6, the data samples of the continuous target postures are corrected,and a posture data set is formed; 7, a posture recognition algorithm is trained by using the obtained training data set. Posture expression in the collected data set can be freely defined artificially, which can support the posture definition operation for the requirements of rehabilitation training of cerebral palsy children by a rehabilitation therapist instead of only supporting limited posture expression of a recognition algorithm technology, therefore a posture expression mode in the system is matched with a posture expression mode understood by the therapist, and the disease pertinenceof the motion sensing game system for rehabilitation training for the cerebral palsy children is enhanced.