In this paper we present an approach for the modeling complex movement sequences. Based on the method of Spatio-Temporal Morphable Models (STMMs) [11] we derive a new hierarchical algorithm that, in a first step, identifies movement elements in the complex movement sequence based on characteristic events, and in a second step quantifies these movement primitives by approximation through linear combinations of learned example movement trajectories. The proposed algorithm is used to segment and to morph sequences of karate movements of different people and different styles.