• Journal of Intelligence and Information Systems
• /
• v.28 no.3
• /
• pp.119-137
• /
• 2022

Recent recommendation system studies apply various deep learning models to represent user and item interactions better. One of the noteworthy studies is ONCF(Outer product-based Neural Collaborative Filtering) which builds a two-dimensional interaction map via outer product and employs CNN (Convolutional Neural Networks) to learn high-order correlations from the map. However, ONCF has limitations in recommendation performance due to the problems with CNN and the absence of side information. ONCF using CNN has an inductive bias problem that causes poor performances for data with a distribution that does not appear in the training data. This paper proposes to employ a Vision Transformer (ViT) instead of the vanilla CNN used in ONCF. The reason is that ViT showed better results than state-of-the-art CNN in many image classification cases. In addition, we propose a new architecture to reflect side information that ONCF did not consider. Unlike previous studies that reflect side information in a neural network using simple input combination methods, this study uses an independent auxiliary classifier to reflect side information more effectively in the recommender system. ONCF used a single latent vector for user and item, but in this study, a channel is constructed using multiple vectors to enable the model to learn more diverse expressions and to obtain an ensemble effect. The experiments showed our deep learning model improved performance in recommendation compared to ONCF.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.17 no.2
• /
• pp.39-57
• /
• 2018

The purpose of this study was to construct an artificial intelligence model that learns and estimates the relationship between vehicle queue length and link travel time in urban areas. The vehicle queue length estimation model is modeled by three models. First of all, classify whether vehicle queue is a link overflow and estimate the vehicle queue length in the link overflow and non-overflow situations. Deep learning model is implemented as Tensorflow. All models are based DNN structure, and network structure which shows minimum error after learning and testing is selected by diversifying hidden layer and node number. The accuracy of the vehicle queue link overflow classification model was 98%, and the error of the vehicle queue estimation model in case of non-overflow and overflow situation was less than 15% and less than 5%, respectively. The average error per link was about 12%. Compared with the detecting data-based method, the error was reduced by about 39%.

• Journal of IKEEE
• /
• v.27 no.4
• /
• pp.486-493
• /
• 2023

In this paper, we propose an automatic hand gesture area extraction and recognition technique using FMCW radar-based point cloud and LSTM. The proposed technique has the following originality compared to existing methods. First, unlike methods that use 2D images as input vectors such as existing range-dopplers, point cloud input vectors in the form of time series are intuitive input data that can recognize movement over time that occurs in front of the radar in the form of a coordinate system. Second, because the size of the input vector is small, the deep learning model used for recognition can also be designed lightly. The implementation process of the proposed technique is as follows. Using the distance, speed, and angle information measured by the FMCW radar, a point cloud containing x, y, z coordinate format and Doppler velocity information is utilized. For the gesture area, the hand gesture area is automatically extracted by identifying the start and end points of the gesture using the Doppler point obtained through speed information. The point cloud in the form of a time series corresponding to the viewpoint of the extracted gesture area is ultimately used for learning and recognition of the LSTM deep learning model used in this paper. To evaluate the objective reliability of the proposed technique, an experiment calculating MAE with other deep learning models and an experiment calculating recognition rate with existing techniques were performed and compared. As a result of the experiment, the MAE value of the time series point cloud input vector + LSTM deep learning model was calculated to be 0.262 and the recognition rate was 97.5%. The lower the MAE and the higher the recognition rate, the better the results, proving the efficiency of the technique proposed in this paper.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.39 no.3
• /
• pp.141-148
• /
• 2021

This paper presents a method to detect zebra-crossing using deep learning which combines SegNet and ResNet. For the blind, a safe crossing system is important to know exactly where the zebra-crossings are. Zebra-crossing detection by deep learning can be a good solution to this problem and robotic vision-based assistive technologies sprung up over the past few years, which focused on specific scene objects using monocular detectors. These traditional methods have achieved significant results with relatively long processing times, and enhanced the zebra-crossing perception to a large extent. However, running all detectors jointly incurs a long latency and becomes computationally prohibitive on wearable embedded systems. In this paper, we propose a model for fast and stable segmentation of zebra-crossing from captured images. The model is improved based on a combination of SegNet and ResNet and consists of three steps. First, the input image is subsampled to extract image features and the convolutional neural network of ResNet is modified to make it the new encoder. Second, through the SegNet original up-sampling network, the abstract features are restored to the original image size. Finally, the method classifies all pixels and calculates the accuracy of each pixel. The experimental results prove the efficiency of the modified semantic segmentation algorithm with a relatively high computing speed.

• Smart Media Journal
• /
• v.9 no.3
• /
• pp.122-129
• /
• 2020

The possibility that an asymptotic coronavirus-19 infected person around the world is not aware of his infection and can spread it to people around him is still a very important issue in that the public is not free from anxiety and fear over the spread of the epidemic. In this paper, the K-means clustering algorithm and deep learning-based crowd aggregation were proposed to determine the possibility of contact with confirmed cases of infectious diseases. As a result of 300 iterations of all input learning images, the PSNR value was 21.51, and the final MAE value for the entire data set was 67.984. This means the average absolute error between observations and the average absolute error of fewer than 4,000 people in each CCTV scene, including the calculation of the distance and infection rate from the confirmed patient and the surrounding persons, the net group of potential patient movements, and the prediction of the infection rate.

• The Journal of the Korea Contents Association
• /
• v.17 no.9
• /
• pp.132-144
• /
• 2017

The fingerprint classification of categorizing fingerprints by classes should be used in order to improve the processing speed and accuracy in a fingerprint recognition system using a large database. The fingerprint classification methods extract features from the fingerprint ridges of a fingerprint and classify the fingerprint using learning and reasoning techniques based on the classes defined according to the flow and shape of the fingerprint ridges. In earlier days, many researches have been conducted using NIST database acquired by pressing or rolling finger against a paper. However, as automated systems using live-scan scanners for fingerprint recognition have become popular, researches using fingerprint images obtained by live-scan scanners, such as fingerprint data provided by FVC, are increasing. And these days the methods of fingerprint classification using Deep Learning have proposed. In this paper, we investigate the trends of fingerprint classification technology and compare the classification performance of the technology. We desire to assist fingerprint classification research with increasing large fingerprint database in improving the performance by mentioning the necessity of fingerprint classification research with consideration for fingerprint images based on live-scan scanners and analyzing fingerprint classification using deep learning.

• International Journal of Computer Science & Network Security
• /
• v.21 no.6
• /
• pp.319-328
• /
• 2021

Parallel administration of numerous drugs increases Drug-Drug Interaction (DDI) because one drug might affect the activity of other drugs. DDI causes negative or positive impacts on therapeutic output. So there is a need to discover DDI to enhance the safety of consuming drugs. Though there are several DDI system exist to predict an interaction but nowadays it becomes impossible to maintain with a large number of biomedical texts which is getting increased rapidly. Mostly the existing DDI system address classification issues, and especially rely on handcrafted features, and some features which are based on particular domain tools. The objective of this paper to predict DDI in a way to avoid adverse effects caused by the consumed drugs, to predict similarities among the drug, Drug pair similarity calculation is performed. The best optimal weight is obtained with the support of KHA. LSTM function with weight obtained from KHA and makes bets prediction of DDI. Our methodology depends on (LSTM-KHA) for the detection of DDI. Similarities among the drugs are measured with the help of drug pair similarity calculation. KHA is used to find the best optimal weight which is used by LSTM to predict DDI. The experimental result was conducted on three kinds of dataset DS1 (CYP), DS2 (NCYP), and DS3 taken from the DrugBank database. To evaluate the performance of proposed work in terms of performance metrics like accuracy, recall, precision, F-measures, AUPR, AUC, and AUROC. Experimental results express that the proposed method outperforms other existing methods for predicting DDI. LSTMKHA produces reasonable performance metrics when compared to the existing DDI prediction model.

• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.488-493
• /
• 2022

In manufacturing sites, bearing fault in eletrically driven motors cause the entire system to shut down. Stopping the operation of this environment causes huge losses in time and money. The reason of this bearing defects can be various factors such as wear due to continuous contact of rotating elements, excessive load addition, and operating environment. In this paper, a motor driving environment is created which is similar to the domestic manufacturing sites. In addition, based on the established environment, we propose a dataset for bearing fault detection by collecting changes in vibration characteristics that vary depending on normal and defective conditions. The sensor used to collect the vibration characteristics is Microphone G.R.A.S. 40PH-10. We used various machine learning models to build a prototype bearing fault detection system trained on the proposed dataset. As the result, based on the deep neural network model, it shows high accuracy performance of 92.3% in the time domain and 98.3% in the frequency domain.

• Journal of the Institute of Convergence Signal Processing
• /
• v.23 no.2
• /
• pp.84-90
• /
• 2022

Machine tool state monitoring is a process that automatically detects the states of machine. In the manufacturing process, the efficiency of machining and the quality of the product are affected by the condition of the tool. Wear and broken tools can cause more serious problems in process performance and lower product quality. Therefore, it is necessary to develop a system to prevent tool wear and damage during the process so that the tool can be replaced in a timely manner. This paper proposes a method for diagnosing five tool states using a deep learning-based hierarchical convolutional neural network to change tools at the right time. The one-dimensional acoustic signal generated when the machine cuts the workpiece is converted into a frequency-based power spectral density two-dimensional image and use as an input for a convolutional neural network. The learning model diagnoses five tool states through three hierarchical steps. The proposed method showed high accuracy compared to the conventional method. In addition, it will be able to be utilized in a smart factory fault diagnosis system that can monitor various machine tools through real-time connecting.

• Journal of Internet Computing and Services
• /
• v.21 no.3
• /
• pp.133-144
• /
• 2020

The number of elderly people with dementia is increasing as fast as the proportion of older people due to aging, which creates a social and economic burden. In particular, dementia care costs, including indirect costs such as increased care costs due to lost caregiver hours and caregivers, have grown exponentially over the years. In order to reduce these costs, it is urgent to introduce a management system to care for dementia patients. Therefore, this study proposes a sensor-based abnormal behavior detection system to manage dementia patients who live alone or in an environment where they cannot always take care of dementia patients. Existing studies were merely evaluating behavior or evaluating normal behavior, and there were studies that perceived behavior by processing images, not data from sensors. In this study, we recognized the limitation of real data collection and used both the auto-encoder, the unsupervised learning model, and the LSTM, the supervised learning model. Autoencoder, an unsupervised learning model, trained normal behavioral data to learn patterns for normal behavior, and LSTM further refined classification by learning behaviors that could be perceived by sensors. The test results show that each model has about 96% and 98% accuracy and is designed to pass the LSTM model when the autoencoder outlier has more than 3%. The system is expected to effectively manage the elderly and dementia patients who live alone and reduce the cost of caring.