http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Wi-Fi RSSI Heat Maps Based Indoor Localization System Using Deep Convolutional Neural Networks
Alwin Poulose,Dong Seog Han 한국방송·미디어공학회 2020 한국방송공학회 학술발표대회 논문집 Vol.2020 No.7
An indoor localization system that uses Wi-Fi RSSI signals for localization gives accurate user position results. The conventional Wi-Fi RSSI signal based localization system uses raw RSSI signals from access points (APs) to estimate the user position. However, the RSSI values of a particular location are usually not stable due to the signal propagation in the indoor environments. To reduce the RSSI signal fluctuations, shadow fading, multipath effects and the blockage of Wi-Fi RSSI signals, we propose a Wi-Fi localization system that utilizes the advantages of Wi-Fi RSSI heat maps. The proposed localization system uses a regression model with deep convolutional neural networks (DCNNs) and gives accurate user position results for indoor localization. The experiment results demonstrate the superior performance of the proposed localization system for indoor localization.
Alwin C. Aguirre 부산외국어대학교 아세안연구원 2017 Suvannabhumi Vol.9 No.1
The paper demonstrates the potential contribution of integrating discursive and affective analytic regimes in framing the study of Southeast Asia. I examine the “emotional possibilities” available to migrants with particular focus on the experience of Filipino domestic helpers in Hong Kong thrown into relief in 2016 by news of maids falling to their deaths while cleaning windows of their employers’ above-ground apartments. First, I situate the study in recent calls for Critical Discourse Studies and Migration Studies to transcend foundational methodologies in their respective fields in order to apprehend formerly disregarded aspects of the human condition, including affect and emotion. I then briefly present the debate in the affective turn in social analysis, which has to do with rethinking the attachment of affect and discourse. My own inquiry is premised on the assertion that emotion is multidimensional. I specifically explore the usefulness of taking emotion as “affective-discursive practice” by focusing on an analysis of the appropriation of the victim role by foreign domestic helper employer groups that could be seen in pertinent news reports of selected online Hong Kong newspapers. In the end, I also emphasize the necessity of reflexivity in projects that take affect as central object of inquiry.
Facial Emotion Recognition Using 3D Face Reconstruction
Alwin Poulose,Chinthala Sreya Reddy,Jung Hwan Kim,Dong Seog Han 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.6
In recent days, autonomous driving systems (ADS) effectively utilize facial emotion recognition (FER) results for safe driving. In FER, the system provides the user emotions such as happy, sad, anger, surprise, disgust, fear, or neutral. These emotions provide helpful information for safe driving and reduce the chances of road accidents. The conventional FER approaches use 2D images as their inputs and classify the user emotions. However, the 2D face images in the conventional FER approaches have limited features for model training. In addition, the features from the 2D face images themselves are not sufficient for accurate emotion classification. To reduce the feature extraction issues in the conventional FER approaches, we propose a 3D face image-based FER approach that uses the 3D face reconstruction technique for converting the 2D face images into 3D face images. The deep convolutional neural networks (DCNNs) used in the proposed FER approach efficiently use the 3D face images as inputs and classify the user emotions with minimum errors. The experiment results show that the proposed 3D face image-based FER approach achieves 99% classification accuracy which is better than the conventional 2D face image-based FER approach.
A Deep Learning Approach for Human Activity Classifier Using Image Data Sets
Alwin Poulose,Mutegeki Ronald,Dong Seog Han 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.2
Human activity recognition (HAR) plays a significant role in health monitoring systems. Most HAR systems use special sensors to identify the user activities. The user activities in the HAR system are divided into basic and compound activities. Sensor-based HAR systems give accurate activity classification results for basic activities. However, when the system has to identify compound activities, the error from the sensors gives poor classification results. To overcome this issue in the HAR systems, we propose a human activity classifier (HAC) using image data instead of sensor data and the proposed HAC shows accurate classification results for compound activities. The proposed HAC system takes the advantages associated with deep convolutional neural networks (DCNNs) to accurately classify the user activities. The experiment results show that the proposed HAC approach achieves 98% model accuracy for activity recognition with 0.012 cross entropy error which is better than other deep learning architectures.