• The Journal of the Convergence on Culture Technology
• /
• v.7 no.4
• /
• pp.675-681
• /
• 2021

With the recent growth of the AI speaker market, the demand for speech synthesis technology that enables natural conversation with users is increasing. Therefore, there is a need for a multi-speaker speech synthesis system that can generate voices of various tones. In order to synthesize natural speech, it is required to train with a large-capacity. high-quality speech DB. However, it is very difficult in terms of recording time and cost to collect a high-quality, large-capacity speech database uttered by many speakers. Therefore, it is necessary to train the speech synthesis system using the speech DB of a very large number of speakers with a small amount of training data for each speaker, and a technique for naturally expressing the tone and rhyme of multiple speakers is required. In this paper, we propose a technology for constructing a speaker encoder by applying the deep learning-based x-vector technique used in speaker recognition technology, and synthesizing a new speaker's tone with a small amount of data through the speaker encoder. In the multi-speaker speech synthesis system, the module for synthesizing mel-spectrogram from input text is composed of Tacotron2, and the vocoder generating synthesized speech consists of WaveNet with mixture of logistic distributions applied. The x-vector extracted from the trained speaker embedding neural networks is added to Tacotron2 as an input to express the desired speaker's tone.

• Animal Bioscience
• /
• v.35 no.12
• /
• pp.1850-1859
• /
• 2022

Objective: The quantitative reverse transcription polymerase chain reaction (qPCR) is the most accurate and reliable technique for analysis of gene expression. Endogenous reference genes (RGs) have been used to normalize qPCR data, although their expression may vary in different tissues and experimental conditions. Verification of the stability of RGs in selected samples is a prerequisite for reliable results. Therefore, we attempted to identify the most stable RGs in the hypothalamic-pituitary-gonadal (HPG) axis in sows. Methods: The cycle threshold values of nine commonly used RGs (18S, HPRT1, GAPDH, RPL4, PPIA, B2M, YWHAZ, ACTB, and SDHA) from HPG axis-related tissues in the domestic sows in the different stages of estrus cycle were analyzed using two RG-finding programs, geNorm and Normfinder, to rank the stability of the pool of RGs. In addition, the effect of the most and least stable RGs was examined by normalization of the target gene, gonadotropin-releasing hormone (GnRH), in the hypothalamus. Results: PPIA, HPRT1, and YWHAZ were the most stable RGs in the HPG axis-related tissues in sows regardless of the stages of estrus cycle. In contrast, traditional RGs, including 18S and ACTB, were found to be the least stable under these experimental conditions. In particular, in the normalization of GnRH expression in the hypothalamus against several stable RGs, PPIA, HPRT1, and YWHAZ, could generate significant (p<0.05) elevation of GnRH in the preovulatory phase compared to the luteal phase, but the traditional RGs with the least stability (18S and ACTB) did not show a significant difference between groups. Conclusion: These results indicate the importance of verifying RG stability prior to commencing research and may contribute to experimental design in the field of animal reproductive physiology as reference data.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.8
• /
• pp.341-354
• /
• 2023

The title of a document is the brief summarization of the document. Readers can easily understand a document if we provide them with its title in their preferred styles and the languages. In this research, we propose a cross-lingual and style-based title generation model using multiple adapters. To train the model, we need a parallel corpus in several languages with different styles. It is quite difficult to construct this kind of parallel corpus; however, a monolingual title generation corpus of the same style can be built easily. Therefore, we apply a zero-shot strategy to generate a title in a different language and with a different style for an input document. A baseline model is Transformer consisting of an encoder and a decoder, pre-trained by several languages. The model is then equipped with multiple adapters for translation, languages, and styles. After the model learns a translation task from parallel corpus, it learns a title generation task from monolingual title generation corpus. When training the model with a task, we only activate an adapter that corresponds to the task. When generating a cross-lingual and style-based title, we only activate adapters that correspond to a target language and a target style. An experimental result shows that our proposed model is only as good as a pipeline model that first translates into a target language and then generates a title. There have been significant changes in natural language generation due to the emergence of large-scale language models. However, research to improve the performance of natural language generation using limited resources and limited data needs to continue. In this regard, this study seeks to explore the significance of such research.

• Journal of Broadcast Engineering
• /
• v.28 no.3
• /
• pp.314-328
• /
• 2023

In this paper, we propose a texture map compression method based on the hierarchical coding method of SHVC to support the scalability function of dynamic mesh compression. The proposed method effectively eliminates the redundancy of multiple-resolution texture maps by downsampling a high-resolution texture map to generate multiple-resolution texture maps and encoding them with SHVC. The dynamic mesh decoder supports the scalability of mesh data by decoding a texture map having an appropriate resolution according to receiver performance and network environment. To evaluate the performance of the proposed method, the proposed method is applied to V-DMC (Video-based Dynamic Mesh Coding) reference software, TMMv1.0, and the performance of the scalable encoder/decoder proposed in this paper and TMMv1.0-based simulcast method is compared. As a result of experiments, the proposed method effectively improves in performance the average of -7.7% and -5.7% in terms of point cloud-based BD-rate (Luma PSNR) in AI and LD conditions compared to the simulcast method, confirming that it is possible to effectively support the texture map scalability of dynamic mesh data through the proposed method.

• Journal of Advanced Navigation Technology
• /
• v.27 no.2
• /
• pp.190-196
• /
• 2023

Recently, following the development of LIDAR technology which can detect distance from the object, the interest for LIDAR based 3D object detection network is getting higher. Previous networks generate inaccurate localization results due to spatial information loss during voxelization and downsampling. In this study, we propose an attention-based convergence method and a camera-LIDAR convergence system to acquire high-level features and high positional accuracy. First, by introducing the attention method into the Voxel-RCNN structure, which is a grid-based 3D object detection network, the multi-scale sparse 3D convolution feature is effectively fused to improve the performance of 3D object detection. Additionally, we propose the late-fusion mechanism for fusing outcomes in 3D object detection network and 2D object detection network to delete false positive. Comparative experiments with existing algorithms are performed using the KITTI data set, which is widely used in the field of autonomous driving. The proposed method showed performance improvement in both 2D object detection on BEV and 3D object detection. In particular, the precision was improved by about 0.54% for the car moderate class compared to Voxel-RCNN.

• SUVANNABHUMI
• /
• v.10 no.2
• /
• pp.181-213
• /
• 2018

How to study Southeast Asia (SEA)? The need to explore and identify methodologies for studying SEA are inherent in its multifaceted subject matter. At a minimum, the region's rich cultural diversity inhibits both the articulation of decisive defining characteristics and the training of scholars who can write with confidence beyond their specialisms. Consequently, the challenges of understanding the region remain and a consensus regarding the most effective approaches to studying its history, identity and future seem quite unlikely. Furthermore, "Area Studies" more generally, has proved to be a less attractive frame of reference for burgeoning scholarly trends. This paper will propose a new tool to help address these challenges. Even though the science of artificial intelligence (AI) is in its infancy, it has already yielded new approaches to many commercial, scientific and humanistic questions. At this point, AI has been used to produce news, generate better smart phones, deliver more entertainment choices, analyze earthquakes and write fiction. The time has come to explore the possibility that AI can be put at the service of the study of SEA. The paper intends to lay out what would be required to develop SEABOT. This instrument might exist as a robot on the web which might be called upon to make the study of SEA both broader and more comprehensive. The discussion will explore the financial resources, ownership and timeline needed to make SEABOT go from an idea to a reality. SEABOT would draw upon artificial neural networks (ANNs) to mine the region's "Big Data", while synthesizing the information to form new and useful perspectives on SEA. Overcoming significant language issues, applying multidisciplinary methods and drawing upon new yields of information should produce new questions and ways to conceptualize SEA. SEABOT could lead to findings which might not otherwise be achieved. SEABOT's work might well produce outcomes which could open up solutions to immediate regional problems, provide ASEAN planners with new resources and make it possible to eventually define and capitalize on SEA's "soft power". That is, new findings should provide the basis for ASEAN diplomats and policy-makers to develop new modalities of cultural diplomacy and improved governance. Last, SEABOT might also open up avenues to tell the SEA story in new distinctive ways. SEABOT is seen as a heuristic device to explore the results which this instrument might yield. More important the discussion will also raise the possibility that an AI-driven perspective on SEA may prove to be even more problematic than it is beneficial.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.28 no.2B
• /
• pp.199-213
• /
• 2008

The goal of this research is to develop and apply the integrational operation method (IOM) for the modeling of the monthly pan evaporation (PE) and the alfalfa reference evapotranspiration ($ET_r$). Since the observed data of the alfalfa $ET_r$ using lysimeter have not been measured for a long time in Republic of Korea, Penman-Monteith (PM) method is used to estimate the observed alfalfa $ET_r$. The IOM consists of the application of the stochastic and neural networks models, respectively. The stochastic model is applied to generate the training dataset for the monthly PE and the alfalfa $ET_r$, and the neural networks models are applied to calculate the observed test dataset reasonably. Among the considered six training patterns, 1,000/PARMA(1,1)/GRNNM-GA training pattern can evaluate the suggested climatic variables very well and also construct the reliable data for the monthly PE and the alfalfa $ET_r$. Uncertainty analysis is used to eliminate the climatic variables of input nodes from 1,000/PARMA(1,1)/GRNNM-GA training pattern. The sensitive and insensitive climatic variables are chosen from the uncertainty analysis of the input nodes. Finally, it can be to model the monthly PE and the alfalfa $ET_r$ simultaneously with the least cost and endeavor using the IOM.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.29 no.5B
• /
• pp.473-482
• /
• 2009

This study numerically discusses wave forces acting on a vertical wall such as breakwaters or revetments, subjected to incident undular or turbulent bores. Due to the complex hydrodynamics of bore, its wave forces have been predicted, mainly through laboratory experiments. Numerical simulations in this paper were carried out by CADMAS-SURF(CDIT, 2001), which is based on Navier-Stokes momentum equations and VOF method (Hirt and Nichols, 1981) for tracking free water surface. Its original source code was also partly revised to generate bore in the numerical water channel. Numerical raw data computed by CADMAS-SURF included great strong spike phenomena that show the abrupt jumps of wave loads. To resolve this undesired noise of raw data, the band-pass filter with the frequency of 5Hz was utilized. The filtered results showed reasonable agreements with the experimental results performed by Matsutomi (1991) and Ramsden (1996). It was confirmed that CADMASSURF can be applied to the design of coastal structures against tsunami bores. In addition, the transformation process and propagation speed of bores in the same 2-d water channel were discussed by the variations of water level for time and space. The numerical results indicated that the propagation speed of bore was changed due to the nonlinear interactions between negative and reflected waves.

• 한국노년학
• /
• v.30 no.4
• /
• pp.1213-1223
• /
• 2010

The purpose of this study is to explore and describe the health and welfare needs experienced by old parents living with disabled adult children, and to help generate research interest and public policy attention on this critical issue. For the purpose of this study, the survey was conducted with older parents who are living with dependent adult children with physical or mental disability. Among collected data for this study, data for 105 older parents were analyzed. The results showed that older parents have suffered with care responsibilities for their disabled adult children as well as special needs resulted from their old age. And older parents have little or even nothing prepared for later life because of lifetime economic, physical and social difficulties related to their disabled children. Also these difficulties had a significant impact on their idea of health and welfare needs in later life. It showed that older parents had mainly concerned and wanted to have the direct cash benefits and medical provisions but hardly recognised the importance of other services such as leisure activities. Preliminary suggestions of this study therefore may be helpful to improve the public policy approach in order to better serve older parents with disabled adult children in the coming aging society.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.25 no.4
• /
• pp.446-454
• /
• 2023

The impact of climate change on agriculture is substantial, especially as global warming is projected to lead to varying temperature and humidity patterns in the future. These changes pose a higher risk for both crops and livestock, exposing them to environmental stressors under altered climatic conditions. Specifically, as temperatures are expected to rise, the risk of heat stress is assessable through the Temperature-Humidity Index (THI), derived from temperature and relative humidity data. This study involved the comparison of THI collected from 10 Korea Meteorological Administration ASOS stations spanning a 60-year period from 1961 to 2020. Moreover, high-resolution temperature and humidity distribution data from 1981 to 2020 were employed to generate high-resolution TH I distributions, analyzing temporal changes. Additionally, the number of days characterized by heat stress, derived from TH I, was compared over different time periods. Generally, TH I showed an upward trend over the past, albeit with varying rates across different locations. As TH I increased, the frequency of heat stress days also rose, indicating potential future cost increases in the livestock industry due to heat-related challenges. The findings emphasize the feasibility of evaluating heat stress risk in livestock using THI and underscore the need for research analyzing THI under future climate change scenarios.