STFT를 사용한 음성 신호 기반의 감정 분류 연구

The human voice has various characteristics, such as, loudness, pitch, speaking rate, etc. This research presents the classification method of human emotions using voice signals transformed using the short-time Fourier transform (STFT). The STFT can know the frequency component at a desired time poi...

Full description

Saved in:
Bibliographic Details
Published in한국생산제조학회지 Vol. 30; no. 5; pp. 366 - 371
Main Authors 신영하(Young-ha Shin), 송규(Kyu Song), 윤찬녕(Chan-nyeong Yun), 조우진(Woo-jin Cho), 박형주(Hyung-joo Park), 장동영(Dong-young Jang)
Format Journal Article
LanguageKorean
Published 한국생산제조학회 01.10.2021
Subjects
기계공학
Speech signals
Emotion recognition
Emotional speech
Speech recognition
Online AccessGet full text
ISSN2508-5093
2508-5107
DOI10.7735/ksmte.2021.30.5.366

Cover

More Information
Summary:The human voice has various characteristics, such as, loudness, pitch, speaking rate, etc. This research presents the classification method of human emotions using voice signals transformed using the short-time Fourier transform (STFT). The STFT can know the frequency component at a desired time point which can be verified using three criteria. Using the 1st criteria, that is, the frequency of the maximum sound intensity (MSI), the emotions can be classified into two groups normal/angry and happy. It is impossible to distinguish between the emotions using the 2nd criteria, which is, the dwell time of the MSI. Using the 3rd criteria, that is, the onset of the MSI, the two groups normal, and angry/happy are identified. Therefore, the 1st and 3rd criteria can be used to classify three emotions. These results can provide valuable insight for future research on the classification of human emotions. KCI Citation Count: 0
ISSN:2508-5093
2508-5107
DOI:10.7735/ksmte.2021.30.5.366