Automatic competency assessment of rhythm performances of ninth-grade and tenth-grade pupils
In this paper, we introduce an approach for automated testing of music competency in rhythm production of ninth-grade and tenth-grade pupils. This work belongs in the larger context of modeling ratings of vocal and instrumental performances. Our approach relies on audio recordings from a specialized mobile application. Rhythmic features were extracted and used to train a machine-learning model which was targeted to approximate human ratings. Using two classes to assess the rhythmic performance, we obtained a mean class accuracy of 0.86.