Supervised and unsupervised feature extraction from a cochlear model for speech recognition
Nathan Intrator, Gary N. Tajchman · 2002
The authors explore the application of a novel classification method that combines supervised and unsupervised training, and compare its performance to various more classical methods. The authors first construct a detailed high dimensional representation of the speech signal using Lyon's cochlear model and then optimally reduce its dimensionality. The resulting low dimensional projection retains the information needed for robust speech recognition.>