---
res:
  bibo_abstract:
  - "Unsupervised blind source separation methods do not require a training phase\r\nand
    thus cannot suffer from a train-test mismatch, which is a common concern in\r\nneural
    network based source separation. The unsupervised techniques can be\r\ncategorized
    in two classes, those building upon the sparsity of speech in the\r\nShort-Time
    Fourier transform domain and those exploiting non-Gaussianity or\r\nnon-stationarity
    of the source signals. In this contribution, spatial mixture\r\nmodels which fall
    in the first category and independent vector analysis (IVA)\r\nas a representative
    of the second category are compared w.r.t. their separation\r\nperformance and
    the performance of a downstream speech recognizer on a\r\nreverberant dataset
    of reasonable size. Furthermore, we introduce a serial\r\nconcatenation of the
    two, where the result of the mixture model serves as\r\ninitialization of IVA,
    which achieves significantly better WER performance than\r\neach algorithm individually
    and even approaches the performance of a much more\r\ncomplex neural network based
    technique.@eng"
  bibo_authorlist:
  - foaf_Person:
      foaf_givenName: Christoph
      foaf_name: Boeddeker, Christoph
      foaf_surname: Boeddeker
      foaf_workInfoHomepage: http://www.librecat.org/personId=40767
  - foaf_Person:
      foaf_givenName: Frederik
      foaf_name: Rautenberg, Frederik
      foaf_surname: Rautenberg
      foaf_workInfoHomepage: http://www.librecat.org/personId=72602
  - foaf_Person:
      foaf_givenName: Reinhold
      foaf_name: Haeb-Umbach, Reinhold
      foaf_surname: Haeb-Umbach
      foaf_workInfoHomepage: http://www.librecat.org/personId=242
  dct_date: 2021^xs_gYear
  dct_language: eng
  dct_title: A Comparison and Combination of Unsupervised Blind Source Separation  Techniques@
...