Discrimination measure determination for word distinction

摘要

The method involves determining a measure for the distinction of a first word from a second word, whereby the first word and the second word include respectively at least one sound section. The second word is presented in form of Hidden-Markov models, whereby each sound section is associated with respectively one Hidden-Markov model which comprises a predetermined amount of conditions. The discrimination measure is determined using an association of the first word on the Hidden-Markov models of the second word. The discrimination measure is preferably performed according to a Viterbi algorithm which determines a path through a Trellis which presents the second word using the conditions of the Hidden-Markov model with respect to a temporal axis.