Max Margin Cosine Loss for Speaker Identification on Short Utterances.
Ruifang JiJunhua CaoXinyuan CaiBo XuPublished in: ISCSLP (2018)
Keyphrases
- speaker identification
- max margin
- multi class
- speech recognition
- gaussian mixture model
- speech signal
- latent variables
- multi label
- broadcast news
- feature extraction
- support vector machine
- action recognition
- natural language
- feature vectors
- image classification
- information retrieval
- hidden markov models
- pattern recognition
- multimedia
- feature selection