Fast i-vector denoising using MAP estimation and a noise distributions database for robust speaker recognition.
Waad Ben KhederDriss MatroufPierre-Michel BousquetJean-François BonastreMoez AjiliPublished in: Comput. Speech Lang. (2017)
Keyphrases
- map estimation
- speaker recognition
- denoising
- noisy images
- noisy environments
- speaker verification
- maximum a posteriori
- maximum a posterior
- gaussian noise
- maximum likelihood
- noise level
- markov random field
- gaussian mixture model
- speaker identification
- image denoising
- noise reduction
- vector quantization
- total variation
- speech recognition
- generative model
- speech signal
- em algorithm
- probabilistic model
- artificial neural networks
- object recognition