메뉴 건너뛰기
.. 내서재 .. 알림
소속 기관/학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색
질문

논문 기본 정보

자료유형
학술대회자료
저자정보
M. Adib Kamali (Kumoh National Institute of Technology) Paul Angelo Oroceo (Kumoh National Institute of Technology) Alexander Pascual (Kumoh National Institute of Technology) Angela Caliwag (Kumoh National Institute of Technology) Wansu Lim (Kumoh National Institute of Technology)
저널정보
대한인간공학회 대한인간공학회 학술대회논문집 2021 대한인간공학회 추계학술대회 및 국제심포지엄
발행연도
2021.11
수록면
127 - 127 (1page)

이용수

표지
📌
연구주제
📖
연구배경
🔬
연구방법
🏆
연구결과
AI에게 요청하기
추천
검색
질문

초록· 키워드

오류제보하기
Objective: This study aims to develop a reliable and efficient keyword spotting (KWS) system suitable for a user interface on an edge device application. In addition, this study also aims to implement KWS on both local (to perform low-level tasks and in condition without internet access) and cloud (to perform high-level tasks). Background: KWS plays a significant role in realizing speech-based user interaction with an edge device. Existing KWS system issues include 1) low accuracy, caused by unrecognized noise signals during training and inference process, 2) high computational complexity, caused by the use of deep learning models with complex architecture, 3) internet connection dependency, caused by running main processes into cloud server, and 4) system responsiveness, can be affected by high latency in data transfer over the internet which is affected by external factors. Method: First, to increase the accuracy, crowd-sourcing techniques is used to obtain training sets which includes variety of audio sample with different accents and voice quality. The quality of the training data is improved using signal augmentation and curation methods. Second, to reduce the computational complexity, the trained model is optimized using a quantization method. In case of increasing accuracy, signal augmentation adds background noise to the training dataset and assists the KWS model to recognize the keyword in a variety of environments. Data curation is used is to process augmented signals which include: collecting, organizing, labeling, cleaning, enhancing, and preserving data for the KWS training process. In addition to signal augmentation and curation, a Mel Frequency Cepstral Coefficient (MFCC) is used to extract the features in an audio signal. The extracted features are then restructure ... 전체 초록 보기

목차

등록된 정보가 없습니다.

참고문헌 (0)

참고문헌 신청

함께 읽어보면 좋을 논문

논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!

이 논문의 저자 정보

이 논문과 함께 이용한 논문

최근 본 자료

전체보기

댓글(0)

0