In order to create an evaluation model based on those feature quantities, this study aims to identify the developmental features of musical expression in young children from the perspective of changing body movement aspects.
In this study, the author examined brand-new feature amounts for machine learning classification and discriminating of the level of musical expression in young children. First, the author provided evidence for the findings of statistical analysis of movement components in early childhood musical expression utilising 3D motion capture and machine learning to assess levels of musical development. In this study, full-body motions were first subjected to an ANOVA. A three-way non-repeated ANOVA was used to quantitatively assess the motion capture data of 3-, 4-, and 5-year-old children in child facilities (n=178). Consequently, there was a statistically significant variation in how the bodily parts moved. Right hand movements, including moving distance and moving average acceleration, showed a significant difference. Second, machine learning techniques such as decision trees, the Sequential Minimum Optimization algorithm (SMO), support vector machines (SVM), and neural networks (multi-layer perceptrons) were used to construct classification models for evaluating the degree of musical development as determined by educators using simultaneously recorded children’s video and related motion capture data. The multi-layer perceptron gave the best confusion matrix results among the various trained classification models, and it showed reasonable classifying precision and utility to support educators in assessing children’s musical development stages. As a result of multilayered perceptron machine learning, the movement of the pelvis has a significant correlation with the degree of musical progression. Its consistency in categorization accuracy suggests that the model may be used to assist educators in determining how well youngsters can express themselves musically.
The author then provided some results of eye tracking on musical expression in a recent study based on the classification and discriminating by machine learning of the developmental degree of musical expression in order to figure out additional feature quantities. In order to research human bodily reaction in relation to cognitive and emotional components, eye-tracking is now often employed. According to the author, eye-tracking data on gazepath, fixations, and saccades provides information that can help us grasp musical expression. Children at child facilities aged 3, 4, and 5 years old (n=118) took part in eye tracking while singing a song while wearing an eye tracker (Tobii3). On the calculated data, quantitative analysis using ANOVA was done. The rise in data, including the frequency and magnitude of saccades as well as the saccade’s moving average velocity, revealed that saccades during early childhood musical expression tended to be greater in major keys than in minor keys. The outcome demonstrated that it was possible to extract useful feature values for machine learning from the computed data of eye movement during musical expression.
Tokoha University, Japan.
Please see the link here: https://stm.bookpi.org/CRLLE-V7/article/view/7540
Keywords: 3D motion capture, musical expression in early childhood, ANOVA, machine learning, classifier, eye tracking.