<![CDATA[This study presents a system for automatic chord recognition from audio recordings using the Pitch Class Profile (PCP) and Support Vector Machine (SVM). PCP was chosen as the primary feature extraction method because it can represent the standard 12 pitch classes in music accurately. SVM was selected as the classification model because of its proven success in previous chord recognition studies, offering high accuracy while remaining efficient. Using the Piano Triads Wavset dataset, which contains 432 triad chords across 12 root notes and three chord types such as major, minor, and diminished, the model was trained and tested in an experiment. The audio data were processed to extract PCP features and normalized before being classified using SVM. Evaluation was carried out using both a default SVM configuration and GridSearchCV optimization. Results show that the optimized model achieved up to 82% accuracy across all chord classes, indicating that the proposed approach can recognize chords reliably even without using deep learning or additional features. The final system also includes real-time prediction by user audio input, using Python and streamlit framework.]]>
Copyrights © 2025
