Fig. 9From: Robustness of ad hoc microphone clustering using speaker embeddings: evaluation under realistic and challenging scenariosHistograms of the DRINR with a the Euclidean distance or b the cosine distance for the second set of scenarios. In this set, the sources are separated by at most three time the critical distance. The DRINRs are computed only for microphones that are part of a source clusterBack to article page