A Universally-Deployable ASR Frontend for Joint Acoustic Echo Cancellation, Speech Enhancement, and Voice Separation.
Tom O'MalleyArun NarayananQuan WangPublished in: CoRR (2022)
Keyphrases
- speech enhancement
- speech sounds
- speech signal
- vocal tract
- automatic speech recognition
- noisy environments
- sound source
- speech recognition
- noisy speech
- mel frequency cepstral coefficients
- linear prediction
- speech synthesis
- acoustic features
- background noise
- speaker identification
- single channel
- spectral analysis
- additive noise
- non stationary
- hidden markov models
- broadcast news
- speaker verification
- image acquisition
- signal to noise ratio
- language model
- multiscale
- noise reduction