Encoding finesse using dynamic movement primitives and low-cost sensors

Dynamic movement primitives use the output of differential equations to form a description of a demonstrated motion task. Nonlinear forcing functions are used as inputs to the differential equations, which permits a varied range of outputs. By parameterizing the forcing function inputs, it is possible to categorize the resulting output motions. In this paper, dynamic movement primitives are used to encode finesse by characterizing the proficiency level of dextrous tasks performed by people. The demonstration data is recorded using low-cost sensors. Experimental results are presented that indicate it is possible to discern the actions of a novice from an experienced practitioner performing a bird re-hang task common to the food processing industry.