K-subspaces and time-delay autoassociators for phoneme recognition | IEEE Conference Publication | IEEE Xplore