End-to-end losses based on speaker basis vectors and all-speaker hard negative mining for speaker verification.
Hee-Soo HeoJee-weon JungIl-Ho YangSung-Hyun YoonHye-jin ShimHa-Jin YuPublished in: CoRR (2019)
Keyphrases
- speaker verification
- end to end
- speaker recognition
- prosodic features
- noisy environments
- basis vectors
- audio visual
- linear combination
- emotion recognition
- multilayer perceptron
- data sets
- basis functions
- face verification
- gaussian mixture model
- high dimensional data
- language identification
- text mining
- image processing
- information retrieval