Variable-rate hierarchical CPC leads to acoustic unit discovery in speech.
Santiago CuervoAdrian LancuckiRicard MarxerPawel RychlikowskiJan ChorowskiPublished in: CoRR (2022)
Keyphrases
- speech recognition systems
- speech recognition
- acoustic features
- emotional speech
- acoustic signal
- speech sounds
- knowledge discovery
- acoustic models
- formant frequencies
- data mining
- prosodic features
- emotion recognition
- speech signal
- pattern discovery
- hidden markov models
- speaker recognition
- speech recognizers
- speaker independent
- endpoint detection
- speaker verification
- source localization
- vocal tract
- speech synthesis
- speaker identification
- broadcast news
- spoken language
- discovery process
- coarse to fine
- hierarchical structure
- visual features
- case study
- information retrieval