Constrained EM estimates for harmonic source separation

Sound source separation based on sinusoid modeling is useful in the recovery of vocals or musical instruments from a single channel record. A constrained iterative method for harmonic source parameter estimation is proposed based on an EM algorithm. The method is intended for harmonic source separation. The problem of coinciding partials, and interference among them in general, is mitigated by constraining the "weak" partials on the stronger ones of the same harmonic source. A useful scheme to determine the weakness of a partial is proposed. The constrained iteration is shown to give more accurate estimates of the sinusoidal parameters which results in good source separation even in most cases of highly overlapping spectra.