MID-Attribute Speaker Generation Using Optimal-Transport-Based Interpolation of Gaussian Mixture Models.
Aya WatanabeShinnosuke TakamichiYuki SaitoDetai XinHiroshi SaruwatariPublished in: ICASSP (2023)
Keyphrases
- gaussian mixture model
- speaker recognition
- speaker identification
- mixture model
- em algorithm
- expectation maximization
- maximum likelihood
- feature vectors
- mel frequency cepstral coefficients
- speaker verification
- covariance matrices
- gaussian mixture
- density estimation
- gaussian distribution
- probability density function
- probability density
- speech recognition
- variational bayes
- speech signal
- closed form
- vector quantization
- non stationary
- unsupervised learning