Recognition of acoustic noise mixtures by combined bottom-up and top-down processing
Jukka Sillanpaa, Aussi Klapuri, Jarno Seppänen, Tuomas I. Virtanen · 2000
In this paper, a system is described for the recognition of mixtures of noise sources in acoustic input signals. The problem is approached by utilizing both bottom-up signal analysis and top-down predictions of higher-level models. The developments are made using musical signals as test material. Validation experiments are presented both for selfgenerated sound mixtures and for real musical recordings. 1