作者: Antoine Deleforge , Florence Forbes , Radu Horaud
DOI: 10.1109/ICASSP.2013.6637612
关键词:
摘要: The sound-source separation and localization (SSL) problems are addressed within a unified formulation. Firstly, mapping between white-noise source locations binaural cues is estimated. Secondly, SSL solved via Bayesian inversion of this in the presence multiple sparse-spectrum emitters (such as speech), noise reverberations. We propose variational EM algorithm which described detail together with initialization convergence issues. Extensive real-data experiments show that method outperforms state-of-the-art both (azimuth elevation).