Single-channel dereverberation for distant-talking speech recognition by combining denoising autoencoder and temporal structure normalization.
Yuma UedaLongbiao WangAtsuhiko KaiXiong XiaoEngsiong ChngHaizhou LiPublished in: ISCSLP (2014)
Keyphrases
- speech recognition
- temporal structure
- single channel
- denoising
- wiener filter
- multi channel
- frequency domain
- hidden markov models
- language model
- image denoising
- speech enhancement
- video data
- speech synthesis
- automatic speech recognition
- spatio temporal
- human activities
- temporal relations
- speech signal
- pattern recognition
- wavelet domain
- noisy images
- noisy environments
- wavelet decomposition
- independent component analysis
- image restoration
- image processing
- vocal tract
- natural images
- prior information
- signal processing
- information retrieval
- machine learning
- noise reduction
- neural network
- image compression
- brain activity