Base of Recorded Equivalences with Noise for Iso-melodic Similarity

This dataset contains the processed sequence features and formal function labels representing the BRENIS corpus. It is designed for evaluating melodic similarity algorithms under high structural and performance variability (e.g., vocal vs. acoustic instrumental covers, different tempo, pitch transpositions). A key feature of this dataset is the inclusion of frame-level formal function labels that classify melodic segments into 'Antecedent' (Melodic Tension), 'Consequent' (Melodic Resolution), and 'Silence' (unvoiced segment) based on a mutually exclusive rule system. The data was generated using the MC-MSA pipeline, which involves vocal separation using BS-Roformer, pitch and energy extraction via RMVPE and RMS, and melodic segmentation via Self-Similarity Matrices (SSMs). This corpus is associated with the paper "Melodic Classification and Mutual-Exclusive Sequence Matching in Cover Song Identification".