Clone-based training with shared-weight encoders extracts robust 12-dimensional speech features, and these features outperform PCA when used as WaveNet conditioning for coding and enhancement.
The shared features are found by encouraging the de- terministic mapping fφ to result in outputs that are maximally similar for all clones, despite their different inputs
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Salient Speech Representations Based on Cloned Networks
Clone-based training with shared-weight encoders extracts robust 12-dimensional speech features, and these features outperform PCA when used as WaveNet conditioning for coding and enhancement.