Noise Robust Acoustic Modeling for Single-Channel Speech Recognition Based on a Stream-Wise Transformer Architecture.
Masakiyo FujimotoHisashi KawaiPublished in: Interspeech (2021)
Keyphrases
- speech recognition
- speech enhancement
- noisy speech
- noisy environments
- single channel
- speech recognition systems
- multi channel
- speech signal
- noise reduction
- speaker independent
- speech recognizers
- signal to noise ratio
- hidden markov models
- automatic speech recognition
- language model
- speech recognizer
- acoustic models
- independent component analysis
- background noise
- speech synthesis
- sound source
- speaker identification
- wiener filter
- frequency domain
- pattern recognition
- acoustic features
- signal processing
- prior information
- vocal tract
- image restoration
- brain activity
- additive noise
- wavelet decomposition