This research explores an assessment of motion learning employed a dynamic time warping (DTW) method for our proposed virtual reality system. This system can capture the skeleton information from a user by the Kinect device to compare to the trainer’s skeleton information. The trainer’s skeleton information is an well-done animation in advance. The learning performance is based on the similarity between the skeleton information of trainer and student. The similarity is computed by the Euclidean distance. To understand the correction of this method, we compare the value of DTW method with the score from six researchers assessing for each experimental sample.