Gaussian Mixture Model Learning for Desired Speech Discrimination Based on Kurtosis of Linear Prediction Residual Signals.
Kohei HayashidaMasato NakayamaTakanobu NishiuraYoichi YamashitaPublished in: IDT/IIMSS/STET (2014)
Keyphrases
- gaussian mixture model
- linear prediction
- speaker recognition
- mixture model
- speaker identification
- speech signal
- image coding
- hidden markov models
- supervised learning
- maximum likelihood
- unsupervised learning
- spectral analysis
- vocal tract
- learning algorithm
- prediction error
- speech recognition
- optical flow
- feature space