A universally-deployable ASR frontend for joint acoustic echo cancellation, speech enhancement, and voice separation.
Thomas R. O'MalleyArun NarayananQuan WangPublished in: INTERSPEECH (2022)
Keyphrases
- speech enhancement
- speech sounds
- speech signal
- automatic speech recognition
- vocal tract
- noisy environments
- sound source
- speech recognition
- noisy speech
- mel frequency cepstral coefficients
- acoustic features
- linear prediction
- background noise
- speech synthesis
- hidden markov models
- speaker identification
- noise reduction
- image acquisition
- broadcast news
- speaker verification
- signal to noise ratio
- spectral analysis
- speaker recognition
- non stationary
- pattern recognition
- information retrieval
- additive noise
- high frequency
- image restoration
- multiscale