• Phonetics and Speech Sciences
• /
• v.10 no.1
• /
• pp.33-38
• /
• 2018

Although deep neural network-based acoustic models have greatly improved the performance of automatic speech recognition (ASR), reverberation still degrades the performance of distant speech recognition in indoor environments. In this paper, we adopt the DenseNet, which has shown great performance results in image classification tasks, to improve the performance of reverberant speech recognition. The DenseNet enables the deep convolutional neural network (CNN) to be effectively trained by concatenating feature maps in each convolutional layer. In addition, we extend the concept of multi-resolution CNN to multi-resolution DenseNet for robust speech recognition in reverberant environments. We evaluate the performance of reverberant speech recognition on the single-channel ASR task in reverberant voice enhancement and recognition benchmark (REVERB) challenge 2014. According to the experimental results, the DenseNet-based acoustic models show better performance than do the conventional CNN-based ones, and the multi-resolution DenseNet provides additional performance improvement.

• Journal of Broadcast Engineering
• /
• v.27 no.6
• /
• pp.840-847
• /
• 2022

In a deep neural network (DNN)-based speech enhancement, using global and local input speech information is closely related to model performance. Recently, a nested U-Net structure that utilizes global and local input data information using multi-scale has bee n proposed. This nested U-Net was also applied to speech enhancement and showed outstanding performance. However, a single skip connection used in nested U-Nets must be modified for the nested structure. In this paper, we propose a multi-level skip connection (MLS) to optimize the performance of the nested U-Net-based speech enhancement algorithm. As a result, the proposed MLS showed excellent performance improvement in various objective evaluation metrics compared to the standard skip connection, which means th at the MLS can optimize the performance of the nested U-Net-based speech enhancement algorithm. In addition, the final proposed m odel showed superior performance compared to other DNN-based speech enhancement models.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.447-450
• /
• 2019

본 논문은 한국어 문장의 감성 분류를 위해 문장의 형태소, 음절, 자소를 입력으로 하는 합성곱층과 DenseNet 을 적용한 Text Multi-channel DenseNet 모델을 제안한다. 맞춤법 오류, 음소나 음절의 축약과 탈락, 은어나 비속어의 남용, 의태어 사용 등 문법적 규칙에 어긋나는 다양한 표현으로 인해 단어 기반 CNN 으로 추출 할 수 없는 특징들을 음절이나 자소에서 추출 할 수 있다. 한국어 감성분석에 형태소 기반 CNN 이 많이 쓰이고 있으나, 본 논문에서 제안한 Text Multi-channel DenseNet 모델은 형태소, 음절, 자소를 동시에 고려하고, DenseNet 에 정보를 밀집 전달하여 문장의 감성 분류의 정확도를 개선하였다. 네이버 영화 리뷰 데이터를 대상으로 실험한 결과 제안 모델은 85.96%의 정확도를 보여 Multi-channel CNN 에 비해 1.45% 더 정확하게 문장의 감성을 분류하였다.

• Journal of Institute of Control, Robotics and Systems
• /
• v.9 no.11
• /
• pp.906-916
• /
• 2003

A multi-robot control algorithm using Petri-Net is proposed for 5vs5 robot soccer. The dynamic environment of robot soccer is modeled by defining the place and transition of each robot and converting it into Petri-Net diagram. Once all the places and transitions of robots are represented by the Petri-Net model, their actions can be chosen according to the roles of robots and position of the ball in soccer game, e.g., offensive, defensive and goalie robot. The proposed modeling method is implemented for soccer robot system. The efficiency and applicability of the proposed multiple-robot control algorithm using Petri-Net are demonstrated through 5vs5 Middle League SimuroSot soccer game.

• Smart Media Journal
• /
• v.9 no.1
• /
• pp.16-22
• /
• 2020

In this study, a neural network method performing both Detection and Classification of diseases and insects in paprika is proposed with Multi-Tasking U-net. Paprika on farms does not have a wide variety of diseases in this study, only two classes such as powdery mildew and mite, which occur relatively frequently are made as the targets. Aiming to this, a U-net is used as a backbone network, and the last layers of the encoder and the decoder of the U-net are utilized for classification and segmentation, respectively. As the result, the encoder of the U-net is shared for both of detection and classification. The training data are composed of 680 normal leaves, 450 mite-damaged leaves, and 370 powdery mildews. The test data are 130 normal leaves, 100 mite-damaged leaves, and 90 powdery mildews. Its test results shows 89% of recognition accuracy.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.6
• /
• pp.697-702
• /
• 2019

We propose a multi-band multi-scale DenseNet with dilated convolution that separates background music signals from broadcast content. Dilated convolution can learn the multi-scale context information represented by spectrogram. In computer simulation experiments, the proposed architecture is shown to improve Signal to Distortion Ratio (SDR) by 0.15 dB and 0.27 dB in 0dB and -10 dB Signal to Noise Ratio (SNR) environments, respectively.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2002.05a
• /
• pp.222-229
• /
• 2002

A value net is a business design that uses digital supply chain concepts to achieve both superior customer satisfaction and company profitability. In order to implement the value net model, information processing and distribution needs to occur in real time. Software agent technology is becoming popular due to the inherent characteristics of autonomy, distributedness and modularity. In this paper, we adopt agent technology to handle all real time decision process, making the value net model a complex multi-agent network of decision makers. For the agents to properly coordinate their respective activities we develop MAVN model, a Web-based multi-agent language grounded in the XML and Java.

• The Journal of Engineering Research
• /
• v.5 no.1
• /
• pp.5-16
• /
• 2004

In the era which computer communication changes from auxiliary means of communication to main means of communication, the messenger system becomes the leader of communication. The multi messenger implemented in this paper used .NET and strong MFC as a language of next generation to put n higher competitiveness than current another messenger systems. In addition, it included face to face chatting function as well as messenger function.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.55 no.3
• /
• pp.181-189
• /
• 2019

Discarding juvenile fishes under girth 16 cm nearly equal to inner perimeter of codend mesh size collected by a cover net method in bottom trawl. The body size of the main five species (mackerel, horse mackerel, sea bream, melon seed and black throat seaperch) was measured for their body length, girth, weight, height and width and analyzed size selectivity. Frequency of penetrating fish as retention in a cover net was less than 40% of total number of juvenile discarding fish. The most of body length or girth of five species were significantly different between in the codend and in the cover net. The 50% selection girth in the cover net ranged 8-11 cm were smaller than those in the codend ranged 9-13 cm by the species respectively. The 50% selection body length was significantly related with the ratio of body height (H) by body width (W) both for in the codend or in the cover net while 50% selection girth was not significantly related with H/W. Furthermore 50% selection fish size by fish species between in the codend and in the cover net was not significantly different both in body length or girth. Therefore, the girth selectivity represented possibly as one unique value regarding fish body shape was considered as more useful method for multi-species catch in trawl.

• Korean Journal of Fisheries and Aquatic Sciences
• /
• v.5 no.1
• /
• pp.11-16
• /
• 1972

During the period from 1966 to 1968, total catches of Spanish mackerel averaged 6,000 to 7,000M/T per annual in Korea, and approximately 70 per cent of this amount was captured by drift nets. In an effort to improve the efficiency of drift nets, some experiments were conducted in 1969 to investigate the selectivity of material and mesh sizes. Seven different mesh sizes (80,85,95,100,105,110 and 115 mm) of both multi- and mono-filament netting were used, and the following results were obtained : 1, The body weight of Spanish mackerel taken with the seven different mesh sizes ranges from 0.5kg to 2.9kg, and the mode of body weight consists of three groups, 1kg ($21%$), 1.3kg($15\%$) and 1.5kg($19\%$). 2. For multi-filament net, 80 to 105mm mesh sizes were suitable to catch those three groups, and a little smalter than these for mono-filament net. 3. For Spanish mackerel only, the mono-filament material proved to have 1.5 times better selectivity than multi-filament : however, the latter proved superior for miscellaneous fish species due to its different size and shape. 4. Multi-filament net showed better selectivity for smaller species than mono-filament. (and mono-filament in general indicated opposite phenomenon.)