Reproduction package for: From musical descriptors to perceived emotion: selective preservation of valence and arousal cues across controlled and real-world music

Reproduction package for the manuscript "From musical descriptors to perceived emotion: selective preservation of valence and arousal cues across controlled and real-world music" (Ashena, Khosrowabadi, and Ghaderi, Shahid Beheshti University). Contents: extract_features.py computes the 42 interpretable audio descriptors with one pipeline applied identically to both corpora (Python, librosa 0.10.1) and includes the DEAM subset-selection procedure; vieillard_features_42.json and deam_features_42.json are the extracted per-clip features (56 Vieillard et al. clips; 800 DEAM clips); selected_items.csv lists the selected DEAM items with their annotations and purity; run_analysis.py reproduces every number, table, and figure in the paper from the feature files (fixed seed); check_selection_consistency.py verifies the subset against the public DEAM annotations. See README.md for usage. Audio is not redistributed: the Vieillard et al. stimuli are available from the original authors, and the DEAM audio and annotations from https://cvml.unige.ch/databases/DEAM/. Licensed under CC BY 4.0: any reuse requires attribution. Please cite this record together with the paper.