Using infra-red motion capture data to measure the bow pressing force in bowed string instruments

A method for estimation of bow pressing force is proposed. Infra-red motion capture data are processed and some descriptors are computed; tilt angle, bow transversal position, bow deformation and hair ribbon deflection. Using the descriptors, machine learning and the force data observe, a model is obtained and trained in order to estimate the bow pressing force. Several processes were followed to get the highest correlation and a lower error. The best results were obtained using a classification process of the data into “force” and “no force” existing, followed by a regression process. The models obtained were used with data of real performance but the result wasn’t good. The movement of the bow during calibration process and real performance data is different, and the calibration bow movement was done by non-musician, so it could be possible that the calibration data doesn’t cover all the possible behave of the bow as in a real performance