Stable learning algorithm for low-distortion blind separation of real speech mixture combining multistage ICA and linear prediction.
Tsuyoki NishikawaHiroshi SaruwatariKiyohiro ShikanoPublished in: NOLISP (2003)
Keyphrases
- blind separation
- multistage
- linear prediction
- speech signal
- blind source separation
- learning algorithm
- prediction error
- independent component analysis
- image coding
- speech recognition
- lossless compression
- non stationary
- source separation
- dynamic programming
- machine learning
- spectral analysis
- independent components
- hidden markov models
- source images
- inter frame
- autoregressive
- subband
- source signals
- principal component analysis
- image segmentation
- optimal policy
- bit rate