Fig. 3From: A large TV dataset for speech and music activity detectionThe length distribution (min) of the TVSM-pseudo datasetBack to article page