Speech Emotion Recognition from Variable-Length Inputs with Triplet Loss Function.
Jian HuangYa LiJianhua TaoZhen LianPublished in: INTERSPEECH (2018)
Keyphrases
- loss function
- variable length
- speech emotion recognition
- fixed length
- pairwise
- support vector
- n gram
- empirical risk
- learning to rank
- boosting framework
- bitstream
- risk minimization
- regularization term
- statistical dependencies
- reproducing kernel hilbert space
- stochastic gradient descent
- solution path
- convex loss functions
- search engine
- boosting algorithms