Comparison of Dimensionality Reduction Techniques on Audio Signals.
Pál Tamás, Dániel T. Várkonyi · ELTE Digital Institutional Repository (EDIT) (Eötvös Loránd University) · 2020
This thesis aims to present a comparison of several combinations of feature extraction and dimensionality reduction methods and their application in audio signal processing. The feature extraction methods to be considered are: STFT, Mel-scaled STFT, high-level features and MFCC. The dimensionality reduction methods are: PCA, t-SNE, Isomap, SOM. The process will be the following, in order: filtering, feature extraction, dimensionality reduction and finally classification and cluster analysis.