A Conformer-Based ASR Frontend for Joint Acoustic Echo Cancellation, Speech Enhancement and Speech Separation.
Tom O'MalleyArun NarayananQuan WangAlex ParkJames WalkerNathan HowardPublished in: ASRU (2021)
Keyphrases
- speech enhancement
- speech signal
- sound source
- noisy speech
- automatic speech recognition
- noisy environments
- speech recognition
- speech sounds
- acoustic features
- vocal tract
- single channel
- background noise
- linear prediction
- noise reduction
- hidden markov models
- speaker identification
- speech synthesis
- spectral analysis
- speaker verification
- non stationary
- spectral subtraction
- mel frequency cepstral coefficients
- speaker recognition
- broadcast news
- speech recognizer
- audio features
- pattern recognition
- additive noise
- information retrieval
- spontaneous speech
- feature extraction