A secure multi-party computation based cohort pruning scheme makes privacy-preserving cohort score normalisation computationally feasible for PLDA-based speaker recognition on binarised voice features.
It is based upon cohort prun- ing using BK speaker representations that allow for efficient computation in the encrypted domain
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Privacy-Preserving Speaker Recognition with Cohort Score Normalisation
A secure multi-party computation based cohort pruning scheme makes privacy-preserving cohort score normalisation computationally feasible for PLDA-based speaker recognition on binarised voice features.