• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.363-363
• /
• 2019

활성화 함수(activation function)는 기계학습(machine learning)의 학습과정에 비선형성을 도입하여 심층적인 학습을 용이하게 하고 예측의 정확도를 높이는 중요한 요소 중 하나이다(Roy et al., 2019). 일반적으로 기계학습에서 사용되고 있는 활성화 함수의 종류에는 계단 함수(step function), 시그모이드 함수(sigmoid 함수), 쌍곡 탄젠트 함수(hyperbolic tangent function), ReLU 함수(Rectified Linear Unit function) 등이 있으며, 예측의 정확도 향상을 위하여 다양한 형태의 활성화 함수가 제시되고 있다. 본 연구에서는 기계학습을 통하여 수위예측 시 정확도 향상을 위하여 Hybrid 활성화 함수를 제안하였다. 연구대상지는 조수간만의 영향을 받는 한강을 대상으로 선정하였으며, 2009년 ~ 2018년까지 10년간의 수문자료를 활용하였다. 수위예측 알고리즘은 Python 내 Tensorflow의 RNN (Recurrent Neural Networks) 모델을 이용하였으며, 강수량, 수위, 조위, 댐 방류량, 하천 유량의 수문자료를 학습시켜 3시간 및 6시간 후의 수위를 예측하였다. 예측정확도 향상을 위하여 입력 데이터는 정규화(Normalization)를 시켰으며, 민감도 분석을 통하여 신경망모델의 은닉층 개수, 학습률의 최적 값을 도출하였다. Hybrid 활성화 함수는 쌍곡 탄젠트 함수와 ReLU 함수를 혼합한 형태로 각각의 가중치($w_1,w_2,w_1+w_2=1$)를 변경하여 정확도를 평가하였다. 그 결과 가중치의 비($w_1/w_2$)에 따라서 예측 결과의 RMSE(Roote Mean Square Error)가 최소가 되고 NSE (Nash-Sutcliffe model Efficiency coefficient)가 최대가 되는 지점과 Peak 수위의 예측정확도가 최대가 되는 지점을 확인할 수 있었다. 본 연구는 현재 Data modeling을 통한 수위예측의 정확도 향상을 위해 기초가 되는 연구이나, 향후 다양한 형태의 활성화 함수를 제안하여 정확도를 향상시킨다면 예측 결과를 통하여 침수예보에 대한 의사결정이 가능할 것으로 기대된다.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1440-1447
• /
• 2019

Currently, the prevalence of autism spectrum disorders in children is reported to be higher and shows various types of disorders. In particular, they are having difficulty in communication due to communication impairment in the area of social communication and need to be improved through training. Thus, this study proposes a method of acquiring voice information through a microphone mounted on a robot designed through preliminary research and using this information to make intelligent motions. An ANN(Artificial Neural Network) was used to classify the speech data into robot motions, and we tried to improve the accuracy by combining the Recurrent Neural Network based on Convolutional Neural Network. The preprocessing of input speech data was analyzed using MFCC(Mel-Frequency Cepstral Coefficient), and the motion of the robot was estimated using various data normalization and neural network optimization techniques. In addition, the designed ANN showed a high accuracy by conducting an experiment comparing the accuracy with the existing architecture and the method of human intervention. In order to design robot motions with higher accuracy in the future and to apply them in the treatment and education environment of children with autism.

• Journal of Korean Society of Transportation
• /
• v.23 no.1
• /
• pp.45-54
• /
• 2005

The main purpose of this study is located at developing Ominibus Monitoring System(OMS) for the purpose of the internal management, which enable to establish job standards, find out matters to be improved, and appreciate for its treatment in a systematic way. It is through developing subjective or objective estimation tool with use of importance, perceived level, and complex index at international airport by each principal service items, The direction of this study came towards for the purpose of developing a metric analysis tool, utilizing the Quantitative Secondary Data, Analysing Perceived Data through airport user surveys, systemizing the data collection-input-analysis process, making data image according to the graph of results, planning service encounter and endowing control attribution, and ensuring competitiveness at the minimal international standards. It is much important to set up a pre-investigation plan on the base of existent foreign literature and actual inspection to international airport. Two tasks have been executed together on the base of this pre-investigation; one is developing subjective estimation standards for departing party, entering party, and airport residence and the other is developing objective standards as complementary methods. The study has processed for the monitoring services at airports regularly and irregularly through developing software system for operating standards after ensuring credibility and feasibility of estimation standards with substantial and statistical way.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.26 no.5D
• /
• pp.821-829
• /
• 2006

Traffic load comprise primary input to pavement design causing pavement damage. therefore it should be proceeded suitable traffic load distribution modeling for pavement design and analysis. Traffic load have been represented by equivalent single axle loads (ESALs) which convert mixed traffic stream into one value for design purposes. But there are some limit to apply ESALs to other roads because it is empirical value developed as part of the original AASHO(American Association of State Highway Officials) road test. There have been many efforts to solve these problems. Several leading country have implemented M-E(Mechanistic-Empirical) design procedures based on mechanical concept. As a result, they established traffic load quantification method using load distribution model known as Axle Load Spectra. This paper details Axle Load Spectra and presents axle load distribution model based on normal mixture distribution function using truck load data collected by WIM system installed in national highway. Axle load spectra and axle load distribution model presented in this paper could be useful for basic data when making traffic load quantification plan for pavement design, overweight vehicle permit plan and pavement maintenance cost plan.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• The KIPS Transactions:PartB
• /
• v.15B no.1
• /
• pp.25-36
• /
• 2008

In this paper, we present a robust hand recognition approach to sudden illumination changes. The proposed approach constructs a background model with respect to hue and hue gradient in HSI color space and extracts a foreground hand region from an input image using the background subtraction method. Eighteen features are defined for a hand pose and multi-class SVM(Support Vector Machine) approach is applied to learn and classify hand poses based on eighteen features. The proposed approach robustly extracts the contour of a hand with variations in illumination by applying the hue gradient into the background subtraction. A hand pose is defined by two Eigen values which are normalized by the size of OBB(Object-Oriented Bounding Box), and sixteen feature values which represent the number of hand contour points included in each subrange of OBB. We compared the RGB-based background subtraction, hue-based background subtraction and the proposed approach with sudden illumination changes and proved the robustness of the proposed approach. In the experiment, we built a hand pose training model from 2,700 sample hand images of six subjects which represent nine numerical numbers from one to nine. Our implementation result shows 92.6% of successful recognition rate for 1,620 hand images with various lighting condition using the training model.

• Journal of IKEEE
• /
• v.27 no.1
• /
• pp.116-121
• /
• 2023

In this paper, we develop a deep learning structure for a complex microbial incubator that applies deep learning prediction result information. The proposed complex microbial incubator consists of pre-processing of complex microbial data, conversion of complex microbial data structure, design of deep learning network, learning of the designed deep learning network, and GUI development applied to the prototype. In the complex microbial data preprocessing, one-hot encoding is performed on the amount of molasses, nutrients, plant extract, salt, etc. required for microbial culture, and the maximum-minimum normalization method for the pH concentration measured as a result of the culture and the number of microbial cells to preprocess the data. In the complex microbial data structure conversion, the preprocessed data is converted into a graph structure by connecting the water temperature and the number of microbial cells, and then expressed as an adjacency matrix and attribute information to be used as input data for a deep learning network. In deep learning network design, complex microbial data is learned by designing a graph convolutional network specialized for graph structures. The designed deep learning network uses a cosine loss function to proceed with learning in the direction of minimizing the error that occurs during learning. GUI development applied to the prototype shows the target pH concentration (3.8 or less) and the number of cells (10⁸ or more) of complex microorganisms in an order suitable for culturing according to the water temperature selected by the user. In order to evaluate the performance of the proposed microbial incubator, the results of experiments conducted by authorized testing institutes showed that the average pH was 3.7 and the number of cells of complex microorganisms was 1.7 × 10⁸. Therefore, the effectiveness of the deep learning structure for the complex microbial incubator applying the deep learning prediction result information proposed in this paper was proven.

• The KIPS Transactions:PartB
• /
• v.13B no.1 s.104
• /
• pp.53-62
• /
• 2006

Object recognition techniques using principal component analysis are disposed to be decreased recognition rate when lighting change of image happens. The purpose of this thesis is to propose an object recognition technique using new PCA analysis method that discriminates an object in database even in the case that the variation of illumination in training images exists. And the object recognition algorithm proposed here represents more enhanced recognition rate using improved k-Nearest Neighbor. In this thesis, we proposed an object recognition algorithm which creates object space by pre-processing and being learned image using histogram equalization and median filter. By spreading histogram of test image using histogram equalization, the effect to change of illumination is reduced. This method is stronger to change of illumination than basic PCA method and normalization, and almost removes effect of illumination, therefore almost maintains constant good recognition rate. And, it compares ingredient projected test image into object space with distance of representative value and recognizes after representative value of each object in model image is made. Each model images is used in recognition unit about some continual input image using improved k-Nearest Neighbor in this thesis because existing method have many errors about distance calculation.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.497-504
• /
• 2020

Constructing speaker embeddings in speaker verification is an important issue. In general, a self-attention mechanism has been applied for speaker embedding encoding. Previous studies focused on training the self-attention in a high-level layer, such as the last pooling layer. In this case, the effect of low-level layers is not well represented in the speaker embedding encoding. In this study, we propose Masked Cross Self-Attentive Encoding (MCSAE) using ResNet. It focuses on training the features of both high-level and low-level layers. Based on multi-layer aggregation, the output features of each residual layer are used for the MCSAE. In the MCSAE, the interdependence of each input features is trained by cross self-attention module. A random masking regularization module is also applied to prevent overfitting problem. The MCSAE enhances the weight of frames representing the speaker information. Then, the output features are concatenated and encoded in the speaker embedding. Therefore, a more informative speaker embedding is encoded by using the MCSAE. The experimental results showed an equal error rate of 2.63 % using the VoxCeleb1 evaluation dataset. It improved performance compared with the previous self-attentive encoding and state-of-the-art methods.

• Korean Journal of Remote Sensing
• /
• v.37 no.3
• /
• pp.379-394
• /
• 2021

Air pollution data in South Korea is provided on a real-time basis by Air Korea stations since 2005. Previous studies have shown the feasibility of gridding air pollution data, but they were confined to a few cities. This paper examines the creation of nationwide gridded maps for PM10 concentration using 333 Air Korea stations with variogram optimization and ordinary kriging. The accuracy of the spatial interpolation was evaluated by various sampling schemes to avoid a too dense or too sparse distribution of the validation points. Using the 114,745 matchups, a four-round blind test was conducted by extracting random validation points for every 365 days in 2019. The overall accuracy was stably high with the MAE of 5.697 ㎍/m³ and the CC of 0.947. Approximately 1,500 cases for high PM10 concentration also showed a result with the MAE of about 12 ㎍/m³ and the CC over 0.87, which means that the proposed method was effective and applicable to various situations. The gridded maps for daily PM10 concentration at the resolution of 0.05° also showed a reasonable spatial distribution, which can be used as an input variable for a gridded prediction of tomorrow's PM10 concentration.