Speaker-Adapted Confidence Measures for ASR Using Deep Bidirectional Recurrent Neural Networks.
Miguel Ángel del AguaAdrià GiménezAlberto SanchísJorge CiveraAlfons JuanPublished in: IEEE ACM Trans. Audio Speech Lang. Process. (2018)
Keyphrases
- recurrent neural networks
- confidence measures
- automatic speech recognition
- ground truth
- speech recognition
- stereo vision
- confidence measure
- neural network
- optical flow
- feed forward
- artificial neural networks
- speech signal
- reservoir computing
- recurrent networks
- hidden markov models
- neural model
- echo state networks
- feedforward neural networks
- broadcast news
- nonlinear dynamic systems
- chaotic time series
- cascade correlation
- stereo images
- stereo matching
- semi supervised
- genetic algorithm