Multichannel Speech Separation with Recurrent Neural Networks from High-Order Ambisonics Recordings.
Lauréline PerotinRomain SerizelEmmanuel VincentAlexandre GuérinPublished in: ICASSP (2018)
Keyphrases
- high order
- recurrent neural networks
- spontaneous speech
- audio visual
- single channel
- higher order
- speech recognition
- neural network
- linear prediction
- feed forward
- low order
- audio recordings
- speech signal
- sound source
- pairwise
- feedforward neural networks
- acoustic features
- echo state networks
- reservoir computing
- recurrent networks
- multi channel
- automatic speech recognition
- nonlinear dynamic systems
- artificial neural networks
- spoken language
- hidden markov models
- cascade correlation
- markov random field
- partial differential equations
- fourth order
- bayesian logistic regression
- tensor analysis
- lower order
- arithmetic coding
- back propagation
- image segmentation