Volleyball Setting Technique Assessment Using a Single Point Sensor

Abstract

The correct technique is one of the main aspects for semi-professional athletes training their volleyball skills. Traditional training and movement assessment though, might not yield the best result to improve the capabilities of a player. Hereby problems or sub-optimal executions of a movement from a technical point of view are often not easily detectable by a coach or without technical support. We investigate the usage of an IMU (inertial measurement unit) combined with an EMG sensor in form of a ‘Myo’ Sensor unit [16], to classify the setting action of a volleyball player to afterwards judge the technical qualities of the movement and suggest improvements like a digital coach. We look into the framework to gather a suitable ground truth and detect the sequence of the actual setting in the datasets. This is then used in combination with a machine learning model to classify the movement. Results show that a subjective direct description of the inaccuracies of the movement as a ground truth is sufficient for this approach. An additional scored function is designed to classify allowed setting actions by the international Volleyball rules [6]. The sequence selection shows optimal results for 54.4% of the samples, 26.6% of the selected sequences show minor displacements. The classification of the setting action shows best results for labels with 2, 3 and 4 classes with an F1-score of 0.74, 0.64 and 0.35, respectively. The classification results are overall reasonable and are especially interesting for the scored function, giving feedback for beginner players. Using the classification model, feedback for the player is created directly through the ground truth labeling.