• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.9
• /
• pp.1183-1189
• /
• 2021

There have been recently an increasing number of people working out at home. However, many of them do not have face-to-face guidance from experts, so they cannot effectively correct their wrong pose. This may lead to strain and injury to those doing home training. To tackle this problem, this paper proposes a video data-based pose classification and correction system for home training. The proposed system classifies poses using the multi-layer perceptron and pose estimation model, and corrects poses based on joint angels estimated. A voting algorithm that considers the results of successive frames is applied to improve the performance of the pose classification model. Multi-layer perceptron model for post classification shows the highest accuracy with 0.9. In addition, it is shown that the proposed voting algorithm improves the accuracy to 0.93.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.26 no.4A
• /
• pp.639-645
• /
• 2006

FRP (Fiber Reinforced Polymer) is an excellent new constructional material in resistibility to corrosion, high intensity, resistibility to fatigue, and plasticity. FBG (Fiber Bragg Grating) sensor is widely used at present as a smart sensor due to lots of advantages such as electric resistance, small-sized material, and high durability. However, with insufficiency of measuring displacement, FBG sensor is used only as a sensor measuring physical properties like strain or temperature. In this study, FRP and FBG sensors are to be hybridized, which could lead to the development of a smart FRP rod. Moreover, developing the estimated model for deflection with neural network method, with the data measured through FBG sensor, could make conquest of a disadvantage of FBG sensor - uniquely used for sensing strain. Artificial neural network is MLP (Multi-layer perceptron), trained within error rate of 0.001. Nonlinear object function and back-propagation algorithm is applied to training and this model is verified with the measured axial displacement through UTM and the estimated numerical values.

• Proceedings of the KIEE Conference
• /
• 1998.11b
• /
• pp.675-677
• /
• 1998

In this paper, we propose the fuzzy inference algorithm with multi-layer structure. MFIS(Multi-layer Fuzzy Inference System) uses PNN(Polynomial Neural networks) structure and the fuzzy inference method. The PNN is the extended structure of the GMDH(Group Method of Data Hendling), and uses several types of polynomials such as linear, quadratic and cubic, as well as the biquadratic polynomial used in the GMDH. In the fuzzy inference method, the simplified and regression polynomial inference methods are used. Here, the regression polynomial inference is based on consequence of fuzzy rules with the polynomial equations such as linear, quadratic and cubic equation. Each node of the MFIS is defined as fuzzy rules and its structure is a kind of neuro-fuzzy structure. We use the training and testing data set to obtain a balance between the approximation and the generalization of process model. Several numerical examples are used to evaluate the performance of the our proposed model.

• Journal of Korea Multimedia Society
• /
• v.24 no.8
• /
• pp.1000-1011
• /
• 2021

In recent years, using Deep Learning methods to apply for medical and biomedical image analysis has seen many advancements. In clinical, using Deep Learning-based approaches for cancer image analysis is one of the key applications for cancer detection and treatment. However, the scarcity and shortage of labeling images make the task of cancer detection and analysis difficult to reach high accuracy. In 2015, the Unet model was introduced and gained much attention from researchers in the field. The success of Unet model is the ability to produce high accuracy with very few input images. Since the development of Unet, there are many variants and modifications of Unet related architecture. This paper proposes a new approach of using Unet++ with pretrained EfficientNet as backbone architecture for breast tumor cell nuclei segmentation and uses the multi-organ transfer learning approach to segment nuclei of breast tumor cells. We attempt to experiment and evaluate the performance of the network on the MonuSeg training dataset and Triple Negative Breast Cancer (TNBC) testing dataset, both are Hematoxylin and Eosin (H & E)-stained images. The results have shown that EfficientUnet++ architecture and the multi-organ transfer learning approach had outperformed other techniques and produced notable accuracy for breast tumor cell nuclei segmentation.

• Phonetics and Speech Sciences
• /
• v.11 no.4
• /
• pp.89-96
• /
• 2019

In this paper, we propose a deep learning architecture that can effectively detect speech segmentation in broadcast contents. We also propose a multi-scale time-dilated layer for learning the temporal changes of feature vectors. We implement several comparison models to verify the performance of proposed model and calculated the frame-by-frame F-score, precision, and recall. Both the proposed model and the comparison model are trained with the same training data, and we train the model using 32 hours of Korean broadcast data which is composed of various genres (drama, news, documentary, and so on). Our proposed model shows the best performance with F-score 91.7% in Korean broadcast data. The British and Spanish broadcast data also show the highest performance with F-score 87.9% and 92.6%. As a result, our proposed model can contribute to the improvement of performance of speech detection by learning the temporal changes of the feature vectors.

• Nuclear Engineering and Technology
• /
• v.50 no.3
• /
• pp.396-410
• /
• 2018

To assist operators to properly assess the current situation of the plant, accurate fault diagnosis methodology should be available and used. A reliable fault diagnosis method is beneficial for the safety of nuclear power plants. The major idea proposed in this work is integrating the merits of different fault diagnosis methodologies to offset their obvious disadvantages and enhance the accuracy and credibility of on-line fault diagnosis. This methodology uses the principle component analysis-based model and multi-flow model to diagnose fault type. To ensure the accuracy of results from the multi-flow model, a mechanical simulation model is implemented to do the quantitative calculation. More significantly, mechanism simulation is implemented to provide training data with fault signatures. Furthermore, one of the distance formulas in similarity measurement-Mahalanobis distance-is applied for on-line failure degree evaluation. The performance of this methodology was evaluated by applying it to the reactor coolant system of a pressurized water reactor. The results of simulation analysis show the effectiveness and accuracy of this methodology, leading to better confidence of it being integrated as a part of the computerized operator support system to assist operators in decision-making.

• Journal of ILASS-Korea
• /
• v.27 no.2
• /
• pp.57-65
• /
• 2022

The purpose of this study is to use machine learning to build a model capable of predicting the flash boiling spray characteristics. In this study, the flash boiling spray was visualized using Shadowgraph visualization technology, and then the spray image was processed with MATLAB to obtain quantitative data of spray characteristics. The experimental conditions were used as input, and the spray characteristics were used as output to train the machine learning model. For the machine learning model, the XGB (extreme gradient boosting) algorithm was used. Finally, the performance of machine learning model was evaluated using R² and RMSE (root mean square error). In order to have enough data to train the machine learning model, this study used 12 injectors with different design parameters, and set various fuel temperatures and ambient pressures, resulting in about 12,000 data. By comparing the performance of the model with different amounts of training data, it was found that the number of training data must reach at least 7,000 before the model can show optimal performance. The model showed different prediction performances for different spray characteristics. Compared with the upstream spray angle and the downstream spray angle, the model had the best prediction performance for the spray tip penetration. In addition, the prediction performance of the model showed a relatively poor trend in the initial stage of injection and the final stage of injection. The model performance is expired to be further enhanced by optimizing the hyper-parameters input into the model.

• Annual Conference of KIPS
• /
• 2022.11a
• /
• pp.308-310
• /
• 2022

With the rise of the Al wave, task-oriented dialogue systems have become one of the popular research directions in academia and industry. Currently, task-oriented dialogue systems mainly adopt pipelined form, which mainly includes natural language understanding, dialogue state decision making, dialogue state tracking and natural language generation. However, pipelining is prone to error propagation, so many task-oriented dialogue systems in the market are only for single-round dialogues. Usually single- domain dialogues have relatively accurate semantic understanding, while they tend to perform poorly on multi-domain, multi-round dialogue datasets. To solve these issues, we developed a paddle-based multi-domain task-oriented Chinese dialogue system. It is based on NEZHA-base pre-training model and CrossWOZ dataset, and uses intention recognition module, dichotomous slot recognition module and NER recognition module to do DST and generate replies based on rules. Experiments show that the dialogue system not only makes good use of the context, but also effectively addresses long-term dependencies. In our approach, the DST of dialogue tracking state is improved, and our DST can identify multiple slotted key-value pairs involved in the discourse, which eliminates the need for manual tagging and thus greatly saves manpower.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.5
• /
• pp.725-733
• /
• 2024

This study aims to develop an artificial intelligence-based model for predicting the causes of chemical accidents, utilizing data on 865 chemical accident situations and cases provided by the Chemical Safety Agency under the Ministry of Environment from January 2014 to January 2024. The research involved training the data using six artificial intelligence models and compared evaluation metrics such as accuracy, precision, recall, and F1 score. Based on 356 chemical accident cases from 2020 to 2024, additional training data sets were applied using chemical accident cause investigations and similar accident prevention measures suggested by the Chemical Safety Agency from 2021 to 2022. Through this process, the Multi-Layer Perceptron (MLP) model showed an accuracy of 0.6590 and a precision of 0.6821. the Multi-Layer Perceptron (MLP) model showed an accuracy of 0.6590 and a precision of 0.6821. The Logistic Regression model improved its accuracy from 0.6647 to 0.7778 and its precision from 0.6790 to 0.7992, confirming that the Logistic Regression model is the most effective for predicting the causes of chemical accidents.

• Journal of Information Processing Systems
• /
• v.15 no.5
• /
• pp.1231-1242
• /
• 2019

This paper presents a two-dimensional attention-based long short-memory (2D-ALSTM) model for stock index prediction, incorporating input attention and temporal attention mechanisms for weighting of important stocks and important time steps, respectively. The proposed model is designed to overcome the long-term dependency, stock selection, and stock volatility delay problems that negatively affect existing models. The 2D-ALSTM model is validated in a comparative experiment involving the two attention-based models multi-input LSTM (MI-LSTM) and dual-stage attention-based recurrent neural network (DARNN), with real stock data being used for training and evaluation. The model achieves superior performance compared to MI-LSTM and DARNN for stock index prediction on a KOSPI100 dataset.