Segregation of Speakers for Speaker Adaptation in TV News Audio.
Ulpu RemesJanne PylkkönenMikko KurimoPublished in: ICASSP (4) (2007)
Keyphrases
- speaker adaptation
- tv news
- speech recognition
- speaker identification
- maximum likelihood
- speaker dependent
- automatic speech recognition
- speaker independent
- speech recognizer
- broadcast news
- news stories
- multimedia
- hidden markov models
- acoustic features
- language model
- scene analysis
- semantic concepts
- speech signal
- visual information
- gaussian mixture model
- pattern recognition
- noisy environments
- speech synthesis
- mel frequency cepstral coefficients
- computer vision
- machine learning
- audio visual
- image collections
- similarity measure
- high level
- metadata