Using an autoencoder with deformable templates to discover features for automated speech recognition

Navdeep Jaitly, Geoffrey E. Hinton