Analysis, perception, and recognition of isolated Korean vowels
Hyun‐Yeol Chung, Shozo Makino, Ken’iti Kido · The Journal of the Acoustical Society of America · 1988
Distribution of the formant frequencies of eight isolated Korean vowels were analyzed and compared to the results of perceptual experiments. Recognition experiments were carried out based on these formant distributions. In the distributions of vowels on the F1-F2 plane, overlappings were found between the vowels [u], [o], [ə],and [Ɯ]. Here. [ɛ] and [e] overlapped heavily as if they were the same vowel. Using perceptual domains defined by the results of listening lests, the dialectal differences and iterspeaker differences were investigated. In a recognition experiment, a speaker-independent recognition with the Bayes decision method was conducted for 1760 isolated vowels spoken by 20 speakers. A recognition rate of 81.2% was obtained with the formant frequencies F1 and F2, while the rate with the LPC cepstrum coefficients was only 76.4%. Normalization for eliminating interspeaker differences showed good results for vowels uttered by speakers of the same dialect. The highest recognition rate of 99.8% was obtained using a supervised learning method.