Despite the high level of automation in industrial production, manual operations still play an important role and contribute significantly to the overall production costs. For the evaluation of these manual processes the ``Methods-Time Measurement'' (MTM) is widely used. This method is applied to real workplaces or mock-ups thereof, while also Virtual Reality (VR) can be used for the representation of such workplaces. However, the evaluation of the workers' performed actions is still done manually, which is a time-consuming and error-prone process. This paper introduces an approach to automatically detect full-body actions of users in VR and consequently derive the appropriate MTM values, without knowledge of a pre-existing workplan. The detection algorithm that was developed is explained in detail and its performance is analyzed through a user study with 30 participants.