Relative transfer function estimation exploiting instantaneous signals and the signal subspace
Maja Taseska, Emanuël A. P. Habets · 2015
Multichannel noise reduction can be achieved without dis torting the desired signals, provided that the relative transfer functions (RTFs) of the sources are known. Many RTF esti mators require periods where only one source is active, which limits their applicability in practice. We propose an RTF esti mator that does not require such periods. A time-varying RTF is computed per time-frequency (TF) bin that corresponds to the dominant source at that bin. We demonstrate that a min imum variance distortionless response (MVDR) filter based on the proposed RTF estimate can extract multiple sources with low distortion. The MVDR filter has maximum degrees of freedom and hence achieves significantly better noise re duction compared to a linearly constrained minimum variance filter that uses a separate RTF for each source.