Original: Maranan,D.,  Alaoui, S.,  Schiphorst, T., Pasquier, P., Subyen,P., Bartram, L. (2014). Designing For Movement: Evaluating Computational Models using LMA Effort Qualities. CHI 2014, One of a CHInd, Toronto, ON, Canada p.991-1000

Authors: Diego Silang Maranan, Sarah Fdili Alaoui, Thecla Schiphorst, Philippe Pasquier,Pattarawut Subyen, Lyn Bartram

Keywords: Movement recognition; Movement analysis; Laban Effort
analysis; Movement analysis; Movement-based interaction.

Summary: In the paper a new framework for recognizing Laban Effort qualities in real-time is explained. It uses a five degree of freedom single accelerometer input stream, Machine Learning mechanisms and Laban Movement Analysis to detect and classify movement quality.

Main Contents:  The paper describes a system called EffortDetect which is able to detect Laban Effort qualities in real-time. In contrast to existing approaches of Effort recognition, which uses position data from movement capturing systems, EffortDetect works with a single 5-DOF accelerometer (X, Y, Z, Pinch, Roll) input stream to determine qualifiers for the movement.

EffortDetect uses Machine Learning mechanisms to classify movements. Therefore, a supervised learning scheme is applied.  In the training phase a professional Laban Movement Analyst recorded some exact Basic Effort Actions (BEAs) with the accelerometer attached to his wrist. The gathered data is then used to train the classifier. In the performance phase other (professional) dancers execute the same BEAs (under supervision of the expert) and EffortDetect computes classifiers for the BEAs and outputs the most likely BEA based on the trained data in real-time.

After the first tests some additional mainly mathematical improvements for weighting different BEA profiles were conducted.

EffortDetect has been successfully used in at least two real dance performances.

The eight BEAs:

 
Ideas for costume evaluation: EffortDetect needed mainly quantitative data for the first evaluations, which is not really helpful for us. But in my opinion it is worth mentioning that they developed a (in their opinion) finish prototype before evaluating with real actors/dancers. Because of time struggle we have now the problem that we have to test pre-alpha versions of our prototype which could lead to meaningless results. I think the evaluation should take place after we decide that the prototypes are ready for testing. The other way around (the prototypes have to be finished when the evaluation starts) is inappropriate.