Verification of a classification prediction method for the development of musical expression in early childhood using a machine learning method based on 3D motion capture data

Abstract

It is widely acknowledged that children take developmental steps in performing musical expression through body movement dynamics. The author presents an approach to verify the method of classification prediction by machine learning using multiple classifiers to evaluate and classify the developmental degree of musical expressions in early childhood. The author addresses this potential solution by showing statistical analysis of full-body 3D motion captured data using such statistical analysis, and applies machine learning measures to predict developmental degrees of musical expression.

In 2016, the author extracted the feature quantities of movement based on the results of the movement analysis of the musical expressions in the MEB program regarding 3-year-old, 4-year-old, and 5-year-old children (n=76). The developmental degree of musical expression was classified into three stages based on video analysis of the musical expression of each child. The 2016 training data as the feature quantity of movement (factors) and the three-stage evaluation (categorical dependent variables) were used as the model training for machine learning. Based on the results in 2017 with 2018, the classification model training results were applied to the acquired data (n=87) in 2019.

The result of sensitivity analysis showed that the moving average acceleration of pelvis, the moving distance of right foot and the moving distance of right hand had a strong influence on the development of musical expression in early childhood. From those analysis results, the appropriateness of the machine learning method using decision trees for classifying the developmental degree of musical expression in early childhood was verified.