Recognition of speech or speech-like sounds using associative memory

摘要

A speech recognition method disclosed herein achieves high accuracy and high speed in speech recognition, In this mr.thod, an acoustic process is effected on an input speech for acquiring a first train of recognition signals and an associative process is carried out using a model of an associative memory called "Associatron" on the first train of recognition signals. Consequently, a second train of recognition signals is acquired through correction of the first train of recognition signals thanks to the basic feature of the associative process. The second train of recognition signals is verified in the light of a dictionary (13) containing words or the like.