• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.11
• /
• pp.1649-1654
• /
• 2021

In this paper, we introduce a regularization of long short-term memory (LSTM) based fall detection system using TensorFlow that can detect falls that can occur in the elderly. Fall detection uses data from a 3-axis acceleration sensor attached to the body of an elderly person and learns about a total of 7 behavior patterns, each of which is a pattern that occurs in daily life, and the remaining 3 are patterns for falls. During training, a normalization process is performed to effectively reduce the loss function, and the normalization performs a maximum-minimum normalization for data and a L2 regularization for the loss function. The optimal regularization conditions of LSTM using several falling parameters obtained from the 3-axis accelerometer is explained. When normalization and regularization rate λ for sum vector magnitude (SVM) are 127 and 0.00015, respectively, the best sensitivity, specificity, and accuracy are 98.4, 94.8, and 96.9%, respectively.

• International Journal of Advanced Culture Technology
• /
• v.6 no.4
• /
• pp.165-171
• /
• 2018

As the artificial intelligence (AI) plays a crucial role in machine translation (MT) which has loomed large as a new translation paradigm, concerns have also arisen if MT can produce a quality product as human translation (HT) can. In fact, several MT experimental studies report cases in which the MT product called post-editing (PE) as equally as HT or often superior ([1],[2],[6]). As motivated from those studies on translation quality between HT and PE, this study set up an experimental situation in which Korean literature was translated into English, comparatively, by 3 translators and 3 post-editors. Afterwards, a group of 3 other Koreans checked for accuracy of HT and PE; a group of 3 English native speakers scored for fluency of HT and PE. The findings are (1) HT took the translation time, at least, twice longer than PE. (2) Both HT and PE produced similar error types, and Mistranslation and Omission were the major errors for accuracy and Grammar for fluency. (3) HT turned to be inferior to PE for both accuracy and fluency.

• Journal of information and communication convergence engineering
• /
• v.20 no.4
• /
• pp.295-302
• /
• 2022

This paper contains the development of a smart power device designed to collect load power data from industrial manufacturing machines, predict future variations in load power data, and detect abnormal data in advance by applying a machine learning-based prediction algorithm. The proposed load power data prediction model is implemented using a Long Short-Term Memory (LSTM) algorithm with high accuracy and relatively low complexity. The Flask and REST API are used to provide prediction results to users in a graphical interface. In addition, we present the results of experiments conducted to evaluate the performance of the proposed approach, which show that our model exhibited the highest accuracy compared with Multilayer Perceptron (MLP), Random Forest (RF), and Support Vector Machine (SVM) models. Moreover, we expect our method's accuracy could be improved by further optimizing the hyperparameter values and training the model for a longer period of time using a larger amount of data.

• ETRI Journal
• /
• v.46 no.2
• /
• pp.205-217
• /
• 2024

We propose a bandwidth prediction approach based on deep learning. The approach is intended to accurately predict the bandwidth of various types of mobile networks. We first use a machine learning technique, namely, the gradient boosting algorithm, to recognize the connected mobile network. Second, we apply a handover detection algorithm based on network recognition to account for vertical handover that causes the bandwidth variance. Third, as the communication performance offered by 3G, 4G, and 5G networks varies, we suggest a bidirectional long short-term memory model with time distribution for bandwidth prediction per network. To increase the prediction accuracy, pretraining and fine-tuning are applied for each type of network. We use a dataset collected at University College Cork for network recognition, handover detection, and bandwidth prediction. The performance evaluation indicates that the handover detection algorithm achieves 88.5% accuracy, and the bandwidth prediction model achieves a high accuracy, with a root-mean-square error of only 2.12%.

• Korean Journal of Cognitive Science
• /
• v.21 no.2
• /
• pp.359-386
• /
• 2010

In order to evaluate the domain-specific model and process-specific model of spatial and nonspatial working memory (WM), this study manipulated the memory load of the delayed response task and examined how the neural correlates of memory load effect was influenced by the stimulus domain (face and location) at the maintenance stage of WM using an event-related fMRI experiment. One or three face stimuli were presented as target stimuli and participants were asked to maintain the face itself (face WM) or the location of face stimuli (spatial WM). The results of recognition judgment accuracy showed no difference between face WM and spatial WM, and showed equivalent memory load effects of both WM. As a result of brian image analysis, memory load effect at maintenance stage showed that inferior, middle, and superior PFC were recruited by both face WM and spatial WM, and showed that VLPFC was the commonly activated area by both WM, supporting functional specialization of PFC by process components of WM. This study provides evidence for process-specific model in which maintenance of WM is associated with VLPFC.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.13 no.8
• /
• pp.3666-3671
• /
• 2012

We evaluate the face detection and recognition of depth image that is obtained by infrared range sensor. and Face recognition was usually focused on accuracy aspect but it is not enough to evaluate the performance in testing for real world application. In this paper, we evaluate the overall performance like accuracy, training, test speed and memory use for the well known face recognition algorithm like PCA, LDA, ICA and SVM. This experiment evaluate the good results of depth and colored depth image compatible with the colored image although the file size of depth and colored depth image is 30%~40% less than the colored image. Whereas, LDA got the good accuracy performance next to the SVM and also shows the good performance in speed and the amount of memory.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.4
• /
• pp.109-116
• /
• 2021

In this paper, we propose a non-intrusive load monitoring(NILM) system which can find the power of each home appliance from the aggregated total power as the activation in the trading market of the distributed resource and the increasing importance of energy management. We transform the amount of appliances' power into a power on-off state by preprocessing. We use LSTM as a model for predicting states based on these data. Accuracy is measured by comparing predicted states with real ones after postprocessing. In this paper, the accuracy is measured with the different number of electronic products, data postprocessing method, and Time step size. When the number of electronic products is 6, the data postprocessing method using the Round function is used, and Time step size is set to 6, the maximum accuracy can be obtained.

• Journal of Internet Computing and Services
• /
• v.23 no.1
• /
• pp.87-93
• /
• 2022

Since 99% of PCs operating in the defense domain use the Windows operating system, detection and response of Window-based malware is very important to keep the defense cyberspace safe. This paper proposes a model capable of detecting malware in a Windows PE (Portable Executable) format. The detection model was designed with an emphasis on rapid update of the training model to efficiently cope with rapidly increasing malware rather than the detection accuracy. Therefore, in order to improve the training speed, the detection model was designed based on a Bidirectional LSTM (Long Short Term Memory) network that can detect malware with minimal sequence data without complicated pre-processing. The experiment was conducted using the EMBER2018 dataset, As a result of training the model with feature sets consisting of three type of sequence data(Byte-Entropy Histogram, Byte Histogram, and String Distribution), accuracy of 90.79% was achieved. Meanwhile, it was confirmed that the training time was shortened to 1/4 compared to the existing detection model, enabling rapid update of the detection model to respond to new types of malware on the surge.

• Journal of Korea Trade
• /
• v.25 no.4
• /
• pp.17-36
• /
• 2021

Purpose - This study provides useful information to stakeholders by forecasting the tramp shipping market, which is a completely competitive market and has a huge fluctuation in freight rates due to low barriers to entry. Moreover, this study provides the most effective parameters for Baltic Dry Index (BDI) prediction and an optimal model by analyzing and comparing deep learning models such as the artificial neural network (ANN), recurrent neural network (RNN), and long short-term memory (LSTM). Design/methodology - This study uses various data models based on big data. The deep learning models considered are specialized for time series models. This study includes three perspectives to verify useful models in time series data by comparing prediction accuracy according to the selection of external variables and comparison between models. Findings - The BDI research reflecting the latest trends since 2015, using weekly data from 1995 to 2019 (25 years), is employed in this study. Additionally, we tried finding the best combination of BDI forecasts through the input of external factors such as supply, demand, raw materials, and economic aspects. Moreover, the combination of various unpredictable external variables and the fundamentals of supply and demand have sought to increase BDI prediction accuracy. Originality/value - Unlike previous studies, BDI forecasts reflect the latest stabilizing trends since 2015. Additionally, we look at the variation of the model's predictive accuracy according to the input of statistically validated variables. Moreover, we want to find the optimal model that minimizes the error value according to the parameter adjustment in the ANN model. Thus, this study helps future shipping stakeholders make decisions through BDI forecasts.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.28 no.4
• /
• pp.469-476
• /
• 2022

This study developed a Recurrent Neural Network model implemented through Long Short-Term Memory (LSTM) that generates long-term tidal level data at Busan Port using tide observation data. The tide levels in Busan Port were predicted by the Korea Hydrographic and Oceanographic Administration (KHOA) using the tide data observed at Busan New Port and Tongyeong as model input data. The model was trained for one month in January 2019, and subsequently, the accuracy was calculated for one year from February 2019 to January 2020. The constructed model showed the highest performance with a correlation coefficient of 0.997 and a root mean squared error of 2.69 cm when the tide time series of Busan New Port and Tongyeong were inputted together. The study's finding reveal that long-term tidal level data prediction of an arbitrary port is possible using the deep learning recurrent neural network model.