An HMM-based Text Segmentation Method Using Variational Bayes Approach and Its Application to LVCSR for Broadcast News.
Takafumi KoshinakaKen-ichi IsoAkitoshi OkumuraPublished in: ICASSP (1) (2005)
Keyphrases
- segmentation method
- broadcast news
- spoken term detection
- variational bayes
- video search
- spoken documents
- news video
- latent variables
- hyperparameters
- bayesian inference
- automatic speech recognition
- active contour model
- segmentation algorithm
- hidden markov models
- active contours
- language model
- image segmentation
- free energy
- video retrieval
- latent dirichlet allocation
- posterior distribution
- information retrieval
- gaussian mixture model
- energy function
- cross validation
- text mining
- closed form
- model selection
- conditional random fields
- video shots
- natural images
- image search
- key frames
- out of vocabulary
- multiscale