Tests of the automatic recognition of words spoken by different voices based on collected groups of model binary spectrograms

A modified model for the automatic recognition of speech based on binary spectrograms in described and tested. The model employs both a 100-term dictionary and a 26-word dictionary. The findings indicate that automatic recognition of words spoken by different voices based on collected groups of standard binary spectrograms using the proposed modified model is successful only with certain types of voices. Thus, the proposed modification does not assure the desired unification of binary spectrographic images of a given word for a certain broad group of speakers. The implications of the findings are discussed.