Motion choreographing based on kinetic energy with accelerometer signal

Motion choreographing is extremely important nowadays. To provide an intuitive control interface and improve the end-user's experience, low-cost accelerometer, which serving as creating and choreographing character animation based on the performers' movements, attracts more and more researchers. This paper proposes a new feature vector using kinetic energy, which can be calculated from acceleration directly. Our method gives the feature vector more accurate physical significance and improves the recognition rate. Furthermore, post-processing for signal refinement can also benefit from it. We build a prototype of motion recognition system to verify our method. Experiments show that the proposed method leads to a higher recognition rate comparing to typical methods on a database of 48 pre-trained signal data with 8 different categories.