A Cross-Entropy-Guided (CEG) Measure for Speech Enhancement Front-End Assessing Performances of Back-End Automatic Speech Recognition.
Li ChaiJun DuChin-Hui LeePublished in: INTERSPEECH (2019)
Keyphrases
- back end
- automatic speech recognition
- cross entropy
- speech signal
- speech enhancement
- noisy environments
- speech recognition
- user friendly
- vocal tract
- data management
- language modeling
- maximum likelihood
- non stationary
- hidden markov models
- log likelihood
- language model
- broadcast news
- linear prediction
- data types
- building blocks
- acoustic features
- noise reduction
- similarity measure
- evaluation metrics
- error function
- scoring function
- neural network
- single channel
- information retrieval
- data analysis
- image processing