http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : 곽진열(Kwak Jin-Yeol) (검색결과 4 건)
- 무료
- 기관 내 무료
- 유료
-
Attention CRNN에 기반한 오디오 이벤트 검출
곽진열(Jin-Yeol Kwak),정용주(Yong-Joo Chung) 한국전자통신학회 2020 한국전자통신학회 논문지 Vol.15 No.3
최근 들어, 오디오 이벤트 검출을 위하여 다양한 딥뉴럴네트워크 기반의 방법들이 제안되어 왔다. 본 연구에서는 베이스라인 CRNN(Convolutional Recurrent Neural Network) 구조에 attention 방식을 도입함으로서 오디오 이벤트 검출의 성능을 향상시키고자 하였다. 베이스라인 CRNN의 입력단에 context gating을 적용하고 출력단에 attention layer을 추가하였다. 또한, 프레임(frame) 단위의 강전사 레이블(strong label)정보 뿐만 아니라 클립(clip) 단위의 약전사 레이블(weakly label) 오디오 데이터를 이용한 학습을 통하여 보다 나은 성능을 이루고자 하였다. DCASE 2018/2019 Challenge Task 4 데이터를 이용한 오디오 이벤트 검출 실험에서 제안된 attention 기반의 CRNN을 통하여 기존의 CRNN 방식에 비해서 최대 66%의 상대적 F-score 향상을 얻을 수 있었다. Recently, various deep neural networks based methods have been proposed for audio event detection. In this study, we improved the performance of audio event detection by adopting an attention approach to a baseline CRNN. We applied context gating at the input of the baseline CRNN and added an attention layer at the output. We improved the performance of the attention based CRNN by using the audio data of strong labels in frame units as well as the data of weak labels in clip levels. In the audio event detection experiments using the audio data from the Task 4 of the DCASE 2018/2019 Challenge, we could obtain maximally a 66% relative increase in the F-score in the proposed attention based CRNN compared with the baseline CRNN.
-
골조공사 공법 대안 선정을 위한 공사기간 및 비용 예측 모델 개발
곽진열(Kwak Jin-Yeol),김경래(Kim Kyung-Rai),신동우(Shin Dong-Woo),차희성(Cha Hee-Sung) 대한건축학회 2011 大韓建築學會論文集 : 構造系 Vol.27 No.7
Selecting the construction methods at the preconstruction phase is extremely important to reduce the cost and duration of structural work. Because it is forming an ever greater part of the construction process and budget. But most construction companies who have conducted apartment projects don’t try to shorten the duration of cycle-time. This problem was verified by survey from construction managers and experts of the structural works. And it was said that they need the system which can compare and analyse the various construction methods of structural work. So, the purpose of this study is to develop a prediction model which is possible to predict the duration and cost of structural work. The prediction model is comparing and analysing the cost and duration of structural work by using the database of method alternatives. The database includes cost data and duration data of various construction methods such as form-work, reinforcing-bar work and concrete work. Also this study verified the effectiveness of the Prediction Model by simulation of apartment construction case. If construction companies were using this prediction model, they are possible to reduce the cost and duration.
-
차분 특징을 이용한 평균-교사 모델의 음향 이벤트 검출 성능 향상
곽진열(Jin-Yeol Kwak),정용주(Yong-Joo Chung) 한국전자통신학회 2021 한국전자통신학회 논문지 Vol.16 No.3
최근 들어, 음향 이벤트 검출을 위하여 CRNN(: Convolutional Recurrent Neural Network) 구조에 기반 한 평균-교사 모델이 대표적으로 사용되고 있다. 평균-교사 모델은 두 개의 병렬 형태의 CRNN을 가진 구조이며, 이들의 출력들의 일치성을 학습 기준으로 사용함으로서 약-전사 레이블(label)과 비-전사 레이블 음향 데이터에 대해서도 효과적인 학습이 가능하다. 본 연구에서는 최신의 평균-교사 모델에 로그-멜 스펙트럼에 대한 차분 특징을 추가적으로 사용함으로서 보다 나은 성능을 이루고자 하였다. DCASE 2018/2019 Challenge Task 4용 학습 및 테스트 데이터를 이용한 음향 이벤트 검출 실험에서 제안된 차분특징을 이용한 평균-교사모델은 기존의 방식에 비해서 최대 8.1%의 상대적 ER(: Error Rate)의 향상을 얻을 수 있었다. Recently, mean-teacher models based on convolutional recurrent neural networks are popularly used in audio event detection. The mean-teacher model is an architecture that consists of two parallel CRNNs and it is possible to train them effectively on the weakly-labelled and unlabeled audio data by using the consistency learning metric at the output of the two neural networks. In this study, we tried to improve the performance of the mean-teacher model by using additional derivative features of the log-mel spectrum. In the audio event detection experiments using the training and test data from the Task 4 of the DCASE 2018/2019 Challenges, we could obtain maximally a 8.1% relative decrease in the ER(Error Rate) in the mean-teacher model using proposed derivative features.
-
곽진열(Jin-Yeol Kwak),정용주(Yong-Joo Chung) 한국정보기술학회 2020 한국정보기술학회논문지 Vol.18 No.6
In this paper, we used derivative features for sound event detection based on deep neural networks. We extracted log-mel-filterbank value for each frame of the audio signal by frequency analysis and its 1st and 2<SUP>nd</SUP> derivative features were extracted for the use by exploiting the correlation between the frames. CRNN which is recently most popular in audio event detection was used as the baseline detector and 64 dimensional log-mel-filterbank outputs and their 1<SUP>st</SUP> and 2<SUP>nd</SUP> derivatives were constructed as independent input feature maps. Global average pooling layer is added at the output of the CRNN to make use of weak and un-label audio data as well as strong label data in the training. In the various training environment, we could observe consistent performance improvement by using the derivative features. From the experimental results using DCASE Challenge 2018/2018 audio data, we could obtain maximally 16.9% relative improvement in F-score.
ScienceON
스콜라연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
