Nonnegative Matrix Partial Co-Factorization for Spectral and Temporal Drum Source Separation
IEEE Journal on Selected Topics in Signal Processing
We address a problem of separating drum sources from monaural mixtures of polyphonic music containing various pitched instruments as well as drums. We consider a spectrogram of music, described by a matrix where each row is associated with intensities of a frequency over time. We employ a joint decomposition to several spectrogram matrices that include two or more column-blocks of the mixture spectrograms (columns of mixture spectrograms are partitioned into 2 or more blocks) and a drumonly
... and a drumonly (drum solo playing) matrix constructed from various drums a priori. To this end, we apply nonnegative matrix partial cofactorization (NMPCF) to these target matrices, in which columnblocks of mixture spectrograms and the drum-only matrix are jointly decomposed, sharing a factor matrix partially, in order to determine common basis vectors that capture the spectral and temporal characteristics of drum sources. Common basis vectors learned by NMPCF capture spectral patterns of drums since they are shared in the decomposition of the drum-only matrix and accommodate temporal patterns of drums because repetitive characteristics are captured by factorizing columnblocks of mixture spectrograms (each of which is associated with different time periods). Experimental results on real-world commercial music signal demonstrate the performance of the proposed method. Index Terms-Blind source separation, music source separation, nonnegative matrix factorization, nonnegative matrix partial co-factorization.