"Peking-AffectNet: A Repertoire-Based Affective Annotation and Acoustic Feature Dataset for Mei School Jingju Singing"

"Peking-AffectNet is a repertoire-based affective annotation and acoustic feature dataset for Mei School Jingju singing. The dataset contains 322 repertoire-level entries, including historical Mei Lanfang record metadata and contemporary Bilibili-derived singing segments, with affective labels organized by a xiqu-informed taxonomy of four primary emotions and six secondary affective categories. For the decoded subset, the dataset provides transparent acoustic descriptors extracted with librosa, including duration, RMS energy, spectral centroid, zero-crossing rate, estimated tempo, and median YIN fundamental frequency. The release is designed for affective computing, cultural AIGC design, intangible cultural heritage analysis, and reproducible studies of culturally grounded vocal emotion. Audio files are not redistributed because of platform and copyright constraints; provenance links, metadata, labels, feature tables, and analysis summaries are provided for auditability."