🤖 AI Summary
Music emotion recognition (MER) faces two key challenges: scarcity of high-quality annotated data and cross-track feature drift. To address these, we introduce Memo2496—the first large-scale expert-annotated instrumental emotion dataset comprising 2,496 tracks with continuous valence-arousal labels—and propose DAMER, a dual-view adaptive framework. Its core contributions are: (1) a dual-stream attention mechanism fusing Mel-spectrogram and cochleagram representations; (2) progressive confidence-based pseudo-labeling, integrating extreme-emotion calibration and consistency filtering with a threshold of 0.25; and (3) style-anchored contrastive memory learning. DAMER further incorporates curriculum-learning-inspired temperature scheduling and Jensen–Shannon divergence to quantify inter-view consistency. Evaluated on Memo2496, 1000songs, and PMEmo, DAMER achieves state-of-the-art arousal classification accuracy—improving by 3.43%, 2.25%, and 0.17%, respectively. Both the Memo2496 dataset and source code are fully open-sourced.
📝 Abstract
Music Emotion Recogniser (MER) research faces challenges due to limited high-quality annotated datasets and difficulties in addressing cross-track feature drift. This work presents two primary contributions to address these issues. Memo2496, a large-scale dataset, offers 2496 instrumental music tracks with continuous valence arousal labels, annotated by 30 certified music specialists. Annotation quality is ensured through calibration with extreme emotion exemplars and a consistency threshold of 0.25, measured by Euclidean distance in the valence arousal space. Furthermore, the Dual-view Adaptive Music Emotion Recogniser (DAMER) is introduced. DAMER integrates three synergistic modules: Dual Stream Attention Fusion (DSAF) facilitates token-level bidirectional interaction between Mel spectrograms and cochleagrams via cross attention mechanisms; Progressive Confidence Labelling (PCL) generates reliable pseudo labels employing curriculum-based temperature scheduling and consistency quantification using Jensen Shannon divergence; and Style Anchored Memory Learning (SAML) maintains a contrastive memory queue to mitigate cross-track feature drift. Extensive experiments on the Memo2496, 1000songs, and PMEmo datasets demonstrate DAMER's state-of-the-art performance, improving arousal dimension accuracy by 3.43%, 2.25%, and 0.17%, respectively. Ablation studies and visualisation analyses validate each module's contribution. Both the dataset and source code are publicly available.