Using Kinect sensors to monitor and provide feedback to patients performing intervention or rehabilitation exercises is an upcoming trend in healthcare. However, the users' motion sequences differ significantly even when doing the same exercise and are not temporally aligned, making the evaluation of the correctness of their movement challenging. We have developed a method to divide the long motion sequence for each exercise into multiple subsequences, each corresponding to the transition of one key pose to another. We also developed a subsequence-based dynamic time warping algorithm that can automatically detect the endpoint of each subsequence with minimum delay, while simultaneously aligning the detected subsequence to the reference subsequence for the exercise. These methods have been integrated into a prototype system for guiding patients at risks for breast-cancer related lymphedema to perform a set of lymphatic exercises in order to promote lymphatic health and reduce the risk of lymphedema. The system can provide relevant feedback to the patient performing an exercise in real time.