List your company

Speech Separation on WSJ0 open speaker set (test)

8.3SDR Improvement (dB)Oracle ensemble
1.1242.9874.856.713Aug 18, 2015Apr 28, 2016Jan 8, 2017Sep 19, 2017Jun 1, 2018Feb 10, 2019Oct 23, 2019

Evaluation results

MethodLinks
Oracle ensembleType=Ensemble2019.108.3
Speech modelType=Single Model2019.108.2
Confidence ensembleType=Ensemble2019.107.6
DCClustering algorithm=k-means, Scope=oracle2015.086.45
DCClustering algorithm=spectral, Scope=oracle2015.086.26
DCClustering algorithm=k-means, Scope=global2015.085.81
Random ensembleType=Ensemble2019.104.8
Environ. modelType=Single Model2019.102.1
Music modelType=Single Model2019.101.4