作者: Katsuhiko Ishiguro , Hiroshi Sawada , Takuma Otsuka , Hiroshi G. Okuno
DOI:
关键词: Computer science 、 Computational auditory scene analysis 、 Cluster analysis 、 Bayesian probability 、 Topic model 、 Pattern recognition 、 Resolution (logic) 、 Acoustic source localization 、 Permutation (music) 、 Unification 、 Artificial intelligence
摘要: Sound source localization and separation with permutation resolution are essential for achieving a computational auditory scene analysis system that can extract useful information from mixture of various sounds. Because existing methods cope separately these problems despite their mutual dependence, the overall result approaches be degraded by any failure in one components. This paper presents unified Bayesian framework to solve simultaneously where regarded as clustering problem. Experimental results confirm our method outperforms state-of-the-art terms quality setups including practical reverberant environments.