E-Motion: Motion Analysis for Musical Performance

📄 Project Abstract

The aim of this project was to find and test appropriate variables in 3D motion that could be used to predict affective state in a stringed musician's performance. Another objective was to examine the extent to which musicians convey affective state in a similar way. Four musicians, three violinists and one viola player, performed the same piece with different affective states, neutral, happiness, sadness, anger and fear. Variables were extracted from this data and the data was tested to find the extent to which cross-musician affective state could be predicted. The project achieved the aims by segmenting the data and calculating variables that could be tested or trained against data from another musician's performance. Some appropriate variables linked to affective state were found and tested against other musician's performance. It was shown that affective state can be predicted across musicians using 3D motion data of their performances. To add to this, the classifier was tested to ...