Integration of whole body movements with virtual reality environments and computer games has many benefits for exercise training and rehabilitation. Such applications can serve as a virtual personal trainer for different exercise therapies. Current examples of this are based on provision of visual feedback to the user via a webcam yet these allow the player to deviate from the desired exercise sequence without direct warning or feedback. This can be solved by tracking body movements using orientation sensors. However, tracking and providing real time feedback for whole body movements for exercise therapies such as Yoga can prove very complex and require the use of a large number of sensors on body segments. In this paper we describe a methodological approach that can facilitate the development of a body movement driven Yoga exercise computer game that can discriminate player performance level with the use of minimum instrumentation.