Key points are not available for this paper at this time.
As an initial step toward a general speech recognition computer program, program was developed to recognize ten English vowels in isolated words of the form |b| -vowel- |t|. The input to the computer was real-time spectral data. The program first determined the rough location of the first two formants. The remaining confusions were resolved by (1) finer determination of the F1 and F2 locations by the use of slope and/or valley information, (2) the use of pitch information, and (3) in certain cases the determination of the position of F3. The over-all score for 21 subjects (11 male and 10 female) was 88%. By the use of duration information, the score was raised to 93%.
Forgie et al. (Sun,) studied this question.