Dynamic compensation of HMM variances using the feature enhancement uncertainty computed from a parametric model of speech distortion.
Li DengJasha DroppoAlex AceroPublished in: IEEE Trans. Speech Audio Process. (2005)
Keyphrases
- parametric models
- speech recognition
- hidden markov models
- speech signal
- multi stream
- automatic speech recognition systems
- dynamic environments
- feature correspondences
- text to speech
- speech recognizer
- image features
- state space
- feature vectors
- video sequences
- automatic speech recognition
- high quality
- keyword spotting
- image processing