Fig. 15From: Generalized independent low-rank matrix analysis using heavy-tailed distributions for blind source separationTempering approach for a GGD- and bt-ILRMA. The first half of the optimization is based on GGD-ILRMA with β=2 and p=1, and the second half is the proposed method with arbitrary parameters, where the NMF source model is retrained using the temporary estimated signal after the first-half optimizationBack to article page