Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Jeongwook Seo has a total of 63 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Task-Aware Multi-Modal Sensing Control for IoT Systems

    by Yeunwoong Kyung, Dogyu Han, Jihoon Sung, Geunho Lee, Jeongwook Seo, Haneul Ko, Jaehyuk Lee

    Future Internet of Things (IoT) systems increasingly rely on multiple sensing modalities to support sensing-intensive tasks such as localization and smart healthcare. However, always collecting all available sensing data can waste energy and communication resources, whereas aggressive sensing reduction may degrade task-level performance. The key challenge is therefore to adapt sensing operations according to the requirements of the target task while avoiding unnecessary sensing effort. To address this challenge, this article presents Task-Aware Multi-Modal Sensing Control (TA-MSC) for IoT systems. TA-MSC treats sensing operation as a task-level-output-driven closed-loop control problem that identifies a low-cost sensing configuration while satisfying a task-level performance requirement. By analyzing task-level outputs from the inference process, TA-MSC evaluates whether the current sensing configuration is sufficient and updates modality-wise sensing levels accordingly. The selected sensing configuration can be realized through practical control dimensions such as modality selection, sensor activation, channel adjustment, and duty-cycle management. We instantiate TA-MSC through a representative multi-modal localization case study, in which sensing energy is evaluated using literature-based sensing and transmission-energy models rather than direct hardware measurements. In the evaluated cross-device analysis, TA-MSC reduces the mean total modeled sensing energy by 52.8% relative to All-Max, while achieving a Top-1 localization accuracy of 96.14% compared with 97.37% for All-Max and a confidence-satisfaction rate of 98.94% compared with 97.68% for All-Max. These results indicate substantial modeled-energy savings with a modest Top-1 accuracy tradeoff, while the confidence-based task condition is satisfied more frequently.

  • Open Access

    Normalized Difference Red-Edge Estimation with Modified DiscoGAN Model

    by Hyeon-Beom Choi, Kwon-Hee Han, Jeongwook Seo

    Vegetation information is important to study the health and productivity of farmlands and forest ecosystems and investigate the types and severity of threats to them. To obtain vegetation information, Normalized Difference Vegetation Index (NDVI) or Normalized Difference Red-Edge (NDRE) is usually used as a single number quantifying vegetation biomass and plant vigor from satellite remote sensing data. Because they indicate different stages of plant growth and focus on different aspects of plant health, the optimal solution for enhancing vegetation information is to use both of them. However, through satellite remote sensing data containing Red, Green, Blue (RGB) and Near-Infrared (NIR) images, we can only calculate the NDVI, not the NDRE that requires the Red-Edge (RE) images. Therefore, in this paper, we propose an NDRE estimation method using the RE images generated from the RGB images by a modified Discover Cross-Domain Relations with Generative Adversarial Networks (DiscoGAN) model. The modified DiscoGAN model was designed by adding some input and hidden layers in generators and discriminators of the original DiscoGAN model to ingest the RGB images with 256 × 256 × 3 dimension and improve the average Normalized Mean Square Error (NMSE) performance. Experimental results showed that the modified DiscoGAN model outperformed the original DiscoGAN model, obtaining the average NMSE of 0.018 between the real RE images and the generated RE images. Moreover, the NDRE estimation method achieved the average NMSE of 0.074 between the real NDRE values and the NDRE estimates.


  • Open Access

    A Negative Emotion Recognition System with Internet of Things-Based Multimodal Biosignal Data

    by Seung-Mi Ham, Hye-Min Lee, Jae-Hyun Lim, Jeongwook Seo

    Previous studies to recognize negative emotions for mental healthcare have used heavy equipment directly attaching electroencephalogram (EEG) electrodes to the head, and they have proposed binary classification methods to identify negative emotions. To tackle this problem, we propose a negative emotion recognition system to collect multimodal biosignal data such as five EEG signals from an EEG headset and heart rate, galvanic skin response, and skin temperature from a smart band for classifying multiple negative emotions. This consists of an Android Internet of Things (IoT) application, a oneM2M-compliant IoT server, and a machine learning server. The Android IoT application uploads the biosignal data to the IoT server. By using the biosignal data stored in the IoT server, the machine learning server recognizes the negative emotions of disgust, fear, and sadness using a multiclass support vector machine (SVM) model with a radial basis function kernel. The experimental results demonstrate that the multimodal biosignal data approach achieves 93% accuracy. Moreover, when considering only data from the smart band, the system achieved 98% accuracy by optimizing the hyperparameters of the multiclass SVM model. Based on these results, we plan to develop a metaverse system that detects and expresses negative emotions in real time.


  • Open Access

    A Negative Emotion Recognition System with IoT-Based Multimodal Biosignal Data

    by Seung-Mi Ham, Hye-Min Lee, Jae-Hyun Lim, Jeongwook Seo

    Previous studies to recognize negative emotions (e.g. disgust, fear, sadness) for mental health care have used heavy equipment directly attaching electroencephalogram (EEG) electrodes to the head, making it difficult to use in daily life, and they have proposed binary classification methods to determine whether negative emotion or not. To tackle this problem, we propose a negative emotion recognition system to collect multimodal biosignal data such as five EEG signals in an EEG headset and heart rate, galvanic skin response, and skin temperature in a smart band for classifying multiple negative emotions. It consists of android Internet of Things (IoT) application, an oneM2M-compliant IoT server, and a machine learning server. The android IoT application upload the biosignal data to the IoT server. By using the biosignal data stored in the IoT server, the machine learning server recognizes the negative emotions of disgust, fear, and sadness using a multi-class support vector machine (SVM) model with a radial basis function kernel (RBF). The experimental results showed that the multi-class SVM model achieved 93% accuracy when considering all the multimodal biosignal data. Moreover, when considering only data in the smart band, it could achieve 98% accuracy by optimizing the hyper-parameter of the RBF kernel.


  • Open Access

    Current Research in Future Information and Communication Engineering 2022

    by Yun Seop Yu, Kwang-Baek Kim, Dongsik Jo, Hee-Cheol Kim, Jeongwook Seo

    The digital revolution has transformed the way we communicate, access information, and interact with technology


  • A Metaverse Emotion Mapping System with an AIoT Facial Expression Recognition Device

    by Hye-Min Lee, Seung-Mi Ham, Hansol Moon, Hye-Min Kwon, Jae-Hyun Rho, Jeongwook Seo

    Recently, there has been a lot of interest in the research of digital twins and digital humans that can connect the virtual and physical worlds through interaction with various platforms, including social and economic activities. However, experiments and applications in the field of digital twins that connect the metaverse and humans are limited. In particular, research on expressing human emotions in the metaverse is also limited. Furthermore, this technology has the potential to be applied in various industries, including gaming, education, healthcare, and marketing, to provide new forms of interaction and experience. Therefore, in this paper, we propose a metaverse emotion mapping system that can inject new facial expression recognition that can be represented in metaverse objects. We implemented facial expression recognition using a deep learning model based on facial expressions captured by a webcam and applied emotion filtering. The filtered emotion data was then transmitted through an IoT server and the Metaverse Inter-working Module to build a system where objects in the metaverse change. This enables users to share emotions with others through metaverse objects, which could enable emotional communication and expression between people in the metaverse.


  • Open Access

    Android IoT Lifelog System and Its Application to Motion Inference

    by Jeongwook Seo

    In social science, health care, digital therapeutics, etc., smartphone data have played important roles to infer users’ daily lives. However, smartphone data collection systems could not be used effectively and widely because they did not exploit any Internet of Things (IoT) standards (e.g., oneM2M) and class labeling methods for machine learning (ML) services. Therefore, in this paper, we propose a novel Android IoT lifelog system complying with oneM2M standards to collect various lifelog data in smartphones and provide two manual and automated class labeling methods for inference of users’ daily lives. The proposed system consists of an Android IoT client application, an oneM2M-compliant IoT server, and an ML server whose high-level functional architecture was carefully designed to be open, accessible, and internationally recognized in accordance with the oneM2M standards. In particular, we explain implementation details of activity diagrams for the Android IoT client application, the primary component of the proposed system. Experimental results verified that this application could work with the oneM2M-compliant IoT server normally and provide corresponding class labels properly. As an application of the proposed system, we also propose motion inference based on three multi-class ML classifiers (i.e., k nearest neighbors, Naive Bayes, and support vector machine) which were created by using only motion and location data (i.e., acceleration force, gyroscope rate of rotation, and speed) and motion class labels (i.e., driving, cycling, running, walking, and stilling). When compared with confusion matrices of the ML classifiers, the k nearest neighbors classifier outperformed the other two overall. Furthermore, we evaluated its output quality by analyzing the receiver operating characteristic (ROC) curves with area under the curve (AUC) values. The AUC values of the ROC curves for all motion classes were more than 0.9, and the macro-average and micro-average ROC curves achieved very high AUC values of 0.96 and 0.99, respectively.


  • Open Access

    Effect of Compressed Sensing Rates and Video Resolutions on a PoseNet Model in an AIoT System

    by Hye-Min Kwon, Jeongwook Seo

    To provide an artificial intelligence service such as pose estimation with a PoseNet model in an Artificial Intelligence of Things (AIoT) system, an Internet of Things (IoT) sensing device sends a large amount of data such as images or videos to an AIoT edge server. This causes serious data traffic problems in IoT networks. To mitigate these problems, we can apply compressed sensing (CS) to the IoT sensing device. However, the AIoT edge server may have poor pose estimation accuracy (i.e., pose score), because it has to recover the CS data received from the IoT sensing device and estimate human pose from the imperfectly recovered data according to CS rates. Therefore, in this paper, we analyze the effect of CS rates (from 100% to 10%) and video resolutions (1280×720, 640×480, 480×360) in the IoT sensing device on the pose score of the PoseNet model in the AIoT edge server. When only considering the meaningful range of CS rates from 100% to 50%, we found that the higher the video resolution, the lower the pose score. At the CS rate of 80%, we could reduce data traffic by 20% despite the degradation in pose score of less than about 0.03 for all video resolutions.


  • LocateUAV: Unmanned Aerial Vehicle Location Estimation via Contextual Analysis in an IoT Environment

    by Naqqash Dilshad, Amin Ullah, Jaeho Kim, Jeongwook Seo

    Object detection supported by unmanned aerial vehicles (UAVs) has generated significant interest in recent years including applications, such as surveillance, search for missing persons, traffic, and disaster management. Location awareness is a challenging task, particularly, the deployment of UAVs in a global positioning system (GPS) restricted environment or GPS sensor failure. To mitigate this problem, we propose LocateUAV, a novel location awareness framework, to detect UAV’s location by processing the data from the visual sensor in real time using a lightweight convolutional neural network (CNN). Assuming that the drone is in an IoT environment, first, the object detection technique is applied to detect the object of interest (OOI), namely, signboard. Subsequently, optical character recognition (OCR) is applied to extract useful contextual information. In the final step, the extracted information is forwarded to the map application programming interface (API) to locate the UAV. We also present a newly created data set for LocateUAV, which comprises challenging scenarios for context analysis. Moreover, we also compress an existing lightweight model up to 45 MB for efficient processing over UAV, which is 19.5% when compared with the size of the original model. Finally, an in-depth comparison of various trained and efficient object detection and OCR techniques is presented to facilitate future research on the development of flex drone that can extract information from the surroundings of a location in a GPS-restricted environment.


  • Open Access

    Data Traffic Reduction with Compressed Sensing in an AIoT System

    by Hye-Min Kwon, Seng-Phil Hong, Mingoo Kang, Jeongwook Seo

    To provide Artificial Intelligence (AI) services such as object detection, Internet of Things (IoT) sensor devices should be able to send a large amount of data such as images and videos. However, this inevitably causes IoT networks to be severely overloaded. In this paper, therefore, we propose a novel oneM2M-compliant Artificial Intelligence of Things (AIoT) system for reducing overall data traffic and offering object detection. It consists of some IoT sensor devices with random sampling functions controlled by a compressed sensing (CS) rate, an IoT edge gateway with CS recovery and domain transform functions related to compressed sensing, and a YOLOv5 deep learning function for object detection, and an IoT server. By analyzing the effects of compressed sensing on data traffic reduction in terms of data rate per IoT sensor device, we showed that the proposed AIoT system can reduce the overall data traffic by changing compressed sensing rates of random sampling functions in IoT sensor devices. In addition, we analyzed the effects of the compressed sensing on YOLOv5 object detection in terms of performance metrics such as recall, precision, mAP50, and mAP, and found that recall slightly decreases but precision remains almost constant even though the compressed sensing rate decreases and that mAP50 and mAP are gradually degraded according to the decreased compressed sensing rate. Consequently, if proper compressed sensing rates are chosen, the proposed AIoT system will reduce the overall data traffic without significant performance degradation of YOLOv5.