AAI_2025_Capstone_Chronicles_Combined

10

Principal Component Analysis Model

Principal component analysis (PCA) is our baseline classical model used to construct a timbre embedding that reduces our engineered features to a lower dimensional linear space. We can then examine which combinations of features contribute most to perceptual diversity. Figure 5 shows the explained variance for the top components, including both per-component variance and the cumulative variance curve.

Figure 5

Explained variance of PCA components applied to engineered timbre descriptors

Note: Early components contain most of the information, suggesting that timbre relationships can be summarized in a low-dimensional space.

342