메뉴 건너뛰기
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)
고객센터 ENG
주제분류

논문 기본 정보

저자정보
(Jeonbuk National University) (Jeonbuk National University)
저널정보
한국멀티미디어학회 멀티미디어학회논문지 멀티미디어학회논문지 제28권 제10호
오류 신고하기

피인용 0

검색

    초록·키워드

    In order to improve the level of voice emotion recognition in games, this study proposes a model that combines a multi-head attention mechanism with a two-dimensional convolutional neural network (2D-CNN) and a Transformer unit. Experiments on the model on the CASIA emotion datasets show that Mel-frequency cepstral coefficients (MFCCs) are one of the key voice emotion features, and the recognition ability is further enhanced when they are combined with other features. Data augmentation techniques were used to significantly improve recognition accuracy, The weighted accuracy reached 90.83%, the precision was 90.76%, and the recall was 90.00%, which were 2.30%, 1.56%, and 3.12% higher than the traditional CNN-LSTM model, respectively. The key is that the attention mechanism can accurately focus on important emotion-related features in voice signals, filter out irrelevant noise, and thus accurately distinguish similar emotions. This model can use the Valence-Arousal emotional space to perceive the players voice emotions in real time, optimize game interactions, enhance game immersion, and provide reliable emotion perception technology.

    최근 본 자료 전체보기

      UCI(KEPA) : I410-151-26-02-094557672