RPCA-DRNN technique for monaural singing voice separation

Abstract In this study, we propose a methodology for separating singing voice from musical accompaniment in monaural mixture. The proposed method uses robust principal component analysis (RPCA), followed by postprocessing, including median filter, morphology, and high-pass to decompose the Subsequently, deep recurrent neural network comprising two jointly optimized parallel-stacked networks (sRNNs) with mask layers trained on limited data computation is applied decomposed components optimize final estimated separated background music further correct misclassified or residual initial separation. experimental results of MIR-1K, ccMixter, MUSDB18 datasets comparison ten existing techniques indicate that achieves competitive performance On MUSDB18, reaches comparable separation quality less training lower computational cost compared other state-of-the-art technique.