{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

Lopez (poster) - Reverberation Suppression based on sparse...

Info icon This preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
Reverberation Suppression based on sparse linear prediction in noisy environments Nicolás LOPEZ 1,2 , Gaël RICHARD 2 , Yves GRENIER 2 , Ivan BOURMEYSTER 1 1 Arkamys - Paris, France / 2 Institut Mines-Télécom ; Télécom ParisTech ; CNRS LTCI - Paris, France [email protected] Proposed Approach A single channel late reverberation and noise suppression method is pre- sented: STFT Estimate Interferences Filter inverse STFT X l X Y y(t) x(t) Φ Z Late reverberation estimated using fre- quency domain linear prediction with sparse constraints Background noise estimated when speech is absent Blind processing is assumed Real time method Late Reverberation Estimation Observation model: X k,n = X early k,n + X late k,n Late reverberation model: ˆ X late k,n = L - 1 X i =0 α k,i X k,n - i - δ = D k,n α k L : model order, δ : delay Estimation with the LASSO : minimize α k || X k,n - D k,n α k || 2 s.t. || α k || 1 λ Sparse prediction vector: α k = [ α k, 0 . . . α k,L - 1 ] T Signal-based dictionary : D k,n = [ X k,n - δ ... X k,n - δ - L +1 ] (1)
Image of page 1
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}