Robust speech recognition using DNN-HMM acoustic model combining noise-aware training with spectral subtraction.
Akihiro AbeKazumasa YamamotoSeiichi NakagawaPublished in: INTERSPEECH (2015)
Keyphrases
- spectral subtraction
- speech signal
- speech recognition
- noisy environments
- hidden markov models
- noisy speech
- automatic speech recognition
- training process
- background noise
- automatic speech recognition systems
- speech enhancement
- non stationary
- vocal tract
- additive noise
- speech synthesis
- speaker identification
- language model
- linear prediction
- hearing aids
- noise model
- training set
- sound source
- pattern recognition
- neural network
- computer vision