• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.938-939
• /
• 2023

코로나-19로 인해 비접촉 서비스의 중요성이 더욱 대두되고 있다. 키보드나 마우스와 같은 기존 입력 장치를 대체하기 위해 사람들은 디지털 기기에서 손을 사용하여 자연스럽고 간단한 입력을 할 수 있게 되었다. 본 논문에서는 미디어파이프(MediaPipe)와 LSTM(Long Short-Term Memory) 딥러닝을 활용하여 손 제스처를 학습하고 비접촉 입력 장치로 구현하는 방법을 제시한다. 이러한 기술은 가상현실(VR; Virtual Reality), 증강현실(AR; Augmented Reality), 메타버스, 키오스크 등에서 활용 가능성이 크다.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.49 no.3
• /
• pp.8-14
• /
• 2012

Improving the speed of image processing is in great demand according to spread of high quality visual media or massive image applications such as 3D TV or movies, AR(Augmented reality). SIMD computer attached to a host computer can accelerate various image processing and massive data operations. MAMS is a multi-access memory system which is, along with multiple processing elements(PEs), adequate for establishing a high performance pipelined SIMD machine. MAMS supports simultaneous access to pq data elements within a horizontal, a vertical, or a block subarray with a constant interval in an arbitrary position in an $M{\times}N$ array of data elements, where the number of memory modules(MMs), m, is a prime number greater than pq. MAMS-PP4 is the first realization of the MAMS architecture, which consists of four PEs in a single chip and five MMs. This paper presents implementation of image processing algorithms and performance analysis for MAMS-PP16 which consists of 16 PEs with 17 MMs in an extension or the prior work, MAMS-PP4. The newly designed MAMS-PP16 has a 64 bit instruction format and application specific instruction set. The author develops a simulator of the MAMS-PP16 system, which implemented algorithms can be executed on. Performance analysis has done with this simulator executing implemented algorithms of processing images. The result of performance analysis verifies consistent response of MAMS-PP16 through the pyramid operation in image processing algorithms comparing with a Pentium-based serial processor. Executing the pyramid operation in MAMS-PP16 results in consistent response of processing time while randomly response time in a serial processor.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.488-495
• /
• 2021

We propose a speech recognition system based on conformer. Conformer is known to be convolution-augmented transformer, which combines transfer model for capturing global information with Convolution Neural Network (CNN) for exploiting local feature effectively. The baseline system is developed to be a transfer-based speech recognition using Long Short-Term Memory (LSTM)-based language model. The proposed system is a system which uses conformer instead of transformer with transformer-based language model. When Electronics and Telecommunications Research Institute (ETRI) speech corpus in AI-Hub is used for our evaluation, the proposed system yields 5.7 % of Character Error Rate (CER) while the baseline system results in 11.8 % of CER. Even though speech corpus is extended into other domain of AI-hub such as NHNdiguest speech corpus, the proposed system makes a robust performance for two domains. Throughout those experiments, we can prove a validation of the proposed system.

• Proceedings of the Ginseng society Conference
• /
• 1998.06a
• /
• pp.289-299
• /
• 1998

In this present study, we investigated the effects of red ginseng extract and its active constituents - Rbl , Re, Rgl on cycloheximide (CXM)-induced amnesia in the passive avoidance task in rats. Red ginseng water extract at 0.05-0.5 g/kg could improve CXM-induced amnesia in rats, Furthermore, the recovery effect of Rbl at 10 mghg administered 30 min before training trial from CXM-induced amnesia was better than those of Rbl administered other time before or after training trial. Rbl at 0.001-0.1 mghg could significantly improve CXM-induced amnesia and at 1 mghg completely augmented, but at 10 mghg its improving effect slightly weakened. Rgl and Re at 0.3-10 mghg could significantly improve CXM-induced amnesia and Rgl at 10 mg/kg completely avgmented. On the other hand, Rbl at 10 mghg could prolong the step through latencies in the training trial. These results suggest the beneficial effect of red ginseng extract on CXM-induced amnesia in rats could mainly due to the contribution of its active constituents - Rbl, Re, Rgl. The improving effect of Rbl on CXM-induced amnesia was best among the three active constituents. But the reduction in the improving effect of Rbl at 10 mg/kg might be due to the decrease in motor activity and attention to the passive avoidance task.

• International Journal of Advanced Culture Technology
• /
• v.12 no.3
• /
• pp.148-158
• /
• 2024

We investigate Autism Spectrum Disorder (ASD), which is typified by deficits in social interaction, repetitive behaviors, limited vocabulary, and cognitive delays. Traditional diagnostic methodologies, reliant on expert evaluations, frequently result in deferred detection and intervention, particularly in South Korea, where there is a dearth of qualified professionals and limited public awareness. In this study, we employ advanced deep learning algorithms to enhance early ASD screening through automated video analysis. Utilizing architectures such as Convolutional Long Short-Term Memory (ConvLSTM), Long-term Recurrent Convolutional Network (LRCN), and Convolutional Neural Networks with Gated Recurrent Units (CNN+GRU), we analyze video data from platforms like YouTube and TikTok to identify stereotypic behaviors (arm flapping, head banging, spinning). Our results indicate that the LRCN model exhibited superior performance with 79.61% accuracy on the augmented platform video dataset and 79.37% on the original SSBD dataset. The ConvLSTM and CNN+GRU models also achieved higher accuracy than the original SSBD dataset. Through this research, we underscore AI's potential in early ASD detection by automating the identification of stereotypic behaviors, thereby enabling timely intervention. We also emphasize the significance of utilizing expanded datasets from social media platform videos in augmenting model accuracy and robustness, thus paving the way for more accessible diagnostic methods.

• Journal of Korea Multimedia Society
• /
• v.9 no.10
• /
• pp.1257-1265
• /
• 2006

As XML is increasingly used, the need of researches which are related with XML in various fields is also augmented. Many XML document management systems have been actively developed especially for the storage, processing and retrieval of XML documents. The BitCube is a three dimensional bitmap index system that could be manipulated efficiently and improves the performance of document retrieval. However, the site of index is increase rapidly, when a new bit is added to the axis. This problem is caused by its three dimensional memory structure with document, path and word. We suggest a path combining strategy of XML documents in this paper to solve the problem of BitCube that mentioned above. To reduce the size of index, our approach combines sibling nodes that has same ancestor paths, and transforms word axis into value axis. The method reduces the size of index, when the system com poses the three dimensional bitmap index. It also improves the speed of retrieving, and takes efficiency in storage space.

• Journal of Information Technology Services
• /
• v.16 no.2
• /
• pp.111-127
• /
• 2017

Computing Performance, sensor, storage, memory, and network costs have been steadily declining, and IoT services have recently become more active. The Internet of Things is linked with Big Data to create new business, and public institutions and corporations are hurry to import Internet of things. As the importance of the Internet of things has increased, the number of devices supporting the IoT has rapidly increased. With the development of the Internet of Things, various types of Internet services are being developed. For this reason, there is an increasing demand for IoT service designers and developers for IoT service case automatic search technology. IoT service designers can avoid duplication with existing services through service case retrieval and developers can save cost and time by combining existing reusable service equipment. This paper proposes IoT-Lite ontology for IoT and Semantic Web service to solve the above-mentioned problems. The existing ontologies for IoT, despite its many advantages, are not widely used by developers because it has not overcome the relatively slow drawbacks of increasing complexity and searching for development. To complement this, this study uses the IoT-Lite ontology introduced by W3C as a model and a semantic web service for automatic system retrieval. 3D camera, GPS, and 9-axis sensor, and IoT-Lite designed by IoT-Lite technique are integrated with the semantic technique and implemented directly.

• Journal of Ginseng Research
• /
• v.36 no.4
• /
• pp.375-382
• /
• 2012

Ginsenoside Rp1 (G-Rp1) is a saponin derivate that provides anti-metastatic activities through inhibition of the NF-${\kappa}B$ pathway. In this study, we examined the effects of G-Rp1 on regulatory T cell (Treg) activation. After treatment of splenocytes with G-Rp1, Tregs exhibited upregulation of IL-10 expression, and along with dendritic cells (DCs), these Tregs showed increased cell number compared to other cell populations. The effect of G-Rp1 on Treg number was augmented in the presence of lipopolysaccharide (LPS), which mimics pathological changes that occur during inflammation. However, depletion of DCs prevented the increase in Treg number in the presence of G-Rp1 and/or LPS. In addition, G-Rp1 promoted the differentiation of the memory types of $CD4^+Foxp3^+CD62L^{low}$ Tregs rather than the generation of new Tregs. In vivo experiments also demonstrated that Tregs and DCs from mice that were fed G-Rp1 for 7 d and then injected with LPS exhibited increased activation compared with those from mice that were injected with LPS alone. Expression of TGF-${\beta}$ and CTLA4 in Tregs was increased, and upregulation of IL-2 and CD80/CD86 expression by DCs affected the suppressive function of Tregs through IL-2 receptors and CTLA4. These data demonstrate that G-Rp1 exerts anti-inflammatory effects by activating Tregs in vitro and in vivo.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2017.10a
• /
• pp.201-204
• /
• 2017

People live in their everyday life busy with studies, part-time jobs, and searching for an ideal job. In their busy routine, they try to find time for themselves and expose their emotions through diverse social network services(SNS). We made a service that we can plant a virtual tree in places we daily visit and go by. You can keep note on the virtual tree and look through the past records. It is a reality based mobile application service that can be used like a diary.In this project we chose the tree as the metaphor and tried to express time passing in a specific place. As our memory is a part of our daily life, we emphasized the meaning of space important.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.351-363
• /
• 2023

Visual structural inspections are an inseparable part of post-earthquake damage assessments. With unmanned aerial vehicles (UAVs) establishing a new frontier in visual inspections, there are major computational challenges in processing the collected massive amounts of high-resolution visual data. We propose twin deep learning models that can provide accurate high-resolution structural components and damage segmentation masks efficiently. The traditional approach to cope with high memory computational demands is to either uniformly downsample the raw images at the price of losing fine local details or cropping smaller parts of the images leading to a loss of global contextual information. Therefore, our twin models comprising Trainable Resizing for high-resolution Segmentation Network (TRS-Net) and DmgFormer approaches the global and local semantics from different perspectives. TRS-Net is a compound, high-resolution segmentation architecture equipped with learnable downsampler and upsampler modules to minimize information loss for optimal performance and efficiency. DmgFormer utilizes a transformer backbone and a convolutional decoder head with skip connections on a grid of crops aiming for high precision learning without downsizing. An augmented inference technique is used to boost performance further and reduce the possible loss of context due to grid cropping. Comprehensive experiments have been performed on the 3D physics-based graphics models (PBGMs) synthetic environments in the QuakeCity dataset. The proposed framework is evaluated using several metrics on three segmentation tasks: component type, component damage state, and global damage (crack, rebar, spalling). The models were developed as part of the 2^nd International Competition for Structural Health Monitoring.