Jointly trains deep clustering embeddings with uPIT plus discriminative fine-tuning to report better speaker-independent separation than either baseline alone on WSJ0-2mix.
As for the separated network, it has only one BLSTM layer with 896 units
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Discriminative Learning for Monaural Speech Separation Using Deep Embedding Features
Jointly trains deep clustering embeddings with uPIT plus discriminative fine-tuning to report better speaker-independent separation than either baseline alone on WSJ0-2mix.