Automated Speech Recognition as a Function of Formal Speech Training and Passage of Time between Template Training and Testing

Abstract

The purpose of the present study was to assess the longevity of templates trained using a speaker-dependent voice recognition system, and to determine whether recognition varies with the degree of formal speech training. Results indicate that vocabulary recognition is fairly stable over a two-month period, and that subjects with formal voice training do not appreciably differ in performance from novice speakers. However, experience with the voice recognition system did result in improved recognition performance for both trained and novice speakers.