eduzhai > Applied Sciences > Engineering >

Consistent Independent Low-Rank Matrix Analysis for Determined Blind Source Separation

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 41 pages

Abstract: Independent low-rank matrix analysis (ILRMA) is the state-of-the-artalgorithm for blind source separation (BSS) in the determined situation (thenumber of microphones is greater than or equal to that of source signals).ILRMA achieves a great separation performance by modeling the powerspectrograms of the source signals via the nonnegative matrix factorization(NMF). Such a highly developed source model can solve the permutation problemof the frequency-domain BSS to a large extent, which is the reason for theexcellence of ILRMA. In this paper, we further improve the separationperformance of ILRMA by additionally considering the general structure ofspectrograms, which is called consistency, and hence we call the proposedmethod Consistent ILRMA. Since a spectrogram is calculated by an overlappingwindow (and a window function induces spectral smearing called main- andside-lobes), the time-frequency bins depend on each other. In other words, thetime-frequency components are related to each other via the uncertaintyprinciple. Such co-occurrence among the spectral components can function as anassistant for solving the permutation problem, which has been demonstrated by arecent study. On the basis of these facts, we propose an algorithm forrealizing Consistent ILRMA by slightly modifying the original algorithm. Itsperformance was extensively evaluated through experiments performed withvarious window lengths and shift lengths. The results indicated severaltendencies of the original and proposed ILRMA that include some topics notfully discussed in the literature. For example, the proposed Consistent ILRMAtends to outperform the original ILRMA when the window length is sufficientlylong compared to the reverberation time of the mixing system.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...