Recognition of Voiced Consonants Using Characteristic Indicators of the Phonemes on the Time Domain
시간 영역 음소 특징 지표를 이용한 유성 자음 인식

Jae Won Lee
2018 KIISE Transactions on Computing Practices  
This paper presents a voiced consonant recognition system using characteristic indicators of the phonemes on the time domain. The objective voiced consonants are 'ㅁ', 'ㄴ', 'ㄹ', 'o' and the recognition can be performed rapidly using the time domain characteristic indicators. First, the input samples constitute the blocks of the similar size and the volatility indicator is calculated for each block. Next, the bulk indicators are calculated for each bulk which is a set of adjacent samples that
more » ... nt samples that have the same sign. The recognition system is implemented by multilayer perceptions using these indicators to generate input features. The experimental results show that the proposed system can achieve 89.3% recognition accuracy, with improvement by 3.6% when compared with an existing system based on formant frequency. The recognition speed of the system is 1.13 msec per syllable.
doi:10.5626/ktcp.2018.24.10.554 fatcat:ixbob72dxvfupjx6qucoqfqf4q