Should We Always Separate?: Switching Between Enhanced and Observed Signals for Overlapping Speech Recognition.
Hiroshi SatoTsubasa OchiaiMarc DelcroixKeisuke KinoshitaTakafumi MoriyaNaoyuki KamoPublished in: CoRR (2021)
Keyphrases
- speech recognition
- noisy environments
- language model
- hidden markov models
- speech processing
- automatic speech recognition
- speech recognizer
- speech signal
- signal processing
- speech synthesis
- pattern recognition
- speech understanding
- speaker identification
- speech recognition technology
- keyword spotting
- speech recognition systems
- speech recognizers
- speech retrieval
- independent component analysis
- speaker independent
- acoustic models
- speech recognition errors
- spectral analysis
- mixture model