Teaching Speech Enhancement Models to Sing: Domain Adaptation from Speech Enhancement to Singing Voice Separation

This dataset accompanies the 2026 paper “Teaching Speech Enhancement Models to Sing: Domain Adaptation from Speech Enhancement to Singing Voice Separation.” It contains all processed test-set audio samples for every model evaluated in the paper. The original mixture and target signals must be obtained from the respective source test sets. Each test audio directory includes CSV files with the corresponding evaluation metrics. A script to reproduce all reported test results (tables and figures) is available in the associated GitHub repository.