• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2021.06a
• /
• pp.183-186
• /
• 2021

인간 자세 추정 연구는 최근 크게 주목 받고 있는 연구 분야이다. 본 연구는 또한, 자기 지도 학습이라고 명명된 딥러닝 기법이 부상하면서 여러 문제가 해결되고 있다. 본 논문에서는, 이러한 문제를 해결하는 딥러닝 기반 인간 자세 추정 방법들을 유형별로 분류해본다. 그리고 각 분류별 설명과 함께 대표적인 방법들을 소개한다. 마지막으로, 결론에서는 본 연구가 앞으로 나아갈 방향에 대한 논의를 제시한다.

• Journal of Broadcast Engineering
• /
• v.25 no.5
• /
• pp.750-757
• /
• 2020

One of the basic tasks for robots to interact with humans is to quickly and accurately grasp human behavior. Therefore, it is necessary to increase the accuracy of human pose recognition when the robot is estimating the human pose and to recognize it as quickly as possible. However, when the human pose is estimated using deep learning, which is a representative method of artificial intelligence technology, recognition accuracy and speed are not satisfied at the same time. Therefore, it is common to select one of a top-down method that has high inference accuracy or a bottom-up method that has high processing speed. In this paper, we propose two methods that complement the disadvantages while including both the advantages of the two methods mentioned above. The first is to perform parallel inference on the server using multi GPU, and the second is to mix bottom-up and One-class Classification. As a result of the experiment, both of the methods presented in this paper showed improvement in speed. If these two methods are applied to the entertainment robot, it is expected that a highly reliable interaction with the audience can be performed.

• Journal of Internet Computing and Services
• /
• v.19 no.6
• /
• pp.41-51
• /
• 2018

Recently, in the field of video surveillance, deep learning based learning method is applied to intelligent video surveillance system, and various events such as crime, fire, and abnormal phenomenon can be robustly detected. However, since occlusion occurs due to the loss of 3d information generated by projecting the 3d real-world in 2d image, it is need to consider the occlusion problem in order to accurately detect the object and to estimate the pose. Therefore, in this paper, we detect moving objects by solving the occlusion problem of object detection process by adding depth information to existing RGB information. Then, using the convolution neural network in the detected region, the positions of the 14 keypoints of the human joint region can be predicted. Finally, in order to solve the self-occlusion problem occurring in the pose estimation process, the method for 3d human pose estimation is described by extending the range of estimation to the 3d space using the predicted result of 2d keypoint and the deep neural network. In the future, the result of 2d and 3d pose estimation of this research can be used as easy data for future human behavior recognition and contribute to the development of industrial technology.

• Journal of Broadcast Engineering
• /
• v.27 no.5
• /
• pp.718-727
• /
• 2022

The Korean domestic cartoon industry has grown explosively by 65% compared to the previous year. Then the market size is expected to exceed KRW 1 trillion. However, excessive work results in health deterioration. Moreover, this working environment makes the production of human resources insufficient, repeating a vicious cycle. Although some tasks require creation activity during cartoon production, there are still a lot of simple repetitive tasks. Therefore, this study aimed to develop a method for creating a character pose through human pose estimation (HPE). The HPE is to detect key points for each joint of a user. The primary role of the proposed method was to make each joint of the character match that of the human. The proposed method enabled us to create the pose of the two-dimensional cartoon character through the results. Furthermore, it was possible to save the static image for one character pose and the video for continuous character pose.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.3
• /
• pp.583-588
• /
• 2024

With the emergence of Brain-Computer Interface (BCI) technology, analyzing mirror neurons has become more feasible. However, evaluating the accuracy of BCI systems that rely on human thoughts poses challenges due to their qualitative nature. To harness the potential of BCI, we propose a new approach to measure accuracy based on the characteristics of mirror neurons in the human brain that are influenced by speech speed, depending on the ultimate goal of movement. In Chapter 2 of this paper, we introduce mirror neurons and provide an explanation of human posture estimation for mirror neurons. In Chapter 3, we present a powerful pose estimation method suitable for real-time dynamic environments using the technique of human posture estimation. Furthermore, we propose a method to analyze the accuracy of BCI using this robotic environment.

• Review of Korean Society for Internet Information
• /
• v.18 no.2
• /
• pp.19-25
• /
• 2017

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.05a
• /
• pp.27-27
• /
• 2004

이 논문에서는 이동 로봇 플랫폼 위에 장착되는 비전 시스템을 이용하여 대상 사람의 얼굴위치를 검출, 사용자 인식을 수행하는 방법론을 제시한다. 본 연구에서 적용대상으로 하는 이동로봇은 실내에서 사용 가능한 홈 서비스 로봇(Hombot-2)으로 인간-로봇 상호작용 (human-robot interaction, HCI)이 중요한 기능 중에 하나이다. 로봇에 장착된 스테레오 비전 카메라에서 획득하게 되는 얼굴 영상은 임의로 움직이는 로봇 작업 반경 밖에 있는 사용자의 특성 상 얼굴 영상이 비교적 작게 얻어지고 정면얼굴에서 벗어난 가변적 얼굴 자세변화를 갖게 된다.(중략)

• Journal of KIISE
• /
• v.42 no.12
• /
• pp.1544-1550
• /
• 2015

Recently, the development of practical wearable robot interfaces has resulted in the emergence of wearable robots such as arm prosthetics or lower-limb exoskeletons. In this paper, we propose a novel method of wrist movement intention estimation based on a regression technique using electromyography of human bio-signals. In daily life, changes in user arm position changes cause decreases in performance by modulating EMG signals. Therefore, we propose an estimation method for robust wrist movement intention for arm position changes, combining several movement intention models based on the regression technique trained by different arm positions. In our experimental results, our method estimates wrist movement intention more accurately than previous methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.3
• /
• pp.558-567
• /
• 2017

In this research, we developed an efficient real-time human exercise pose estimation system using multiple Kinects. The main objective of this system is to measure and recognize the user's posture (such as knee curl or lunge) more accurately by employing Kinects on the front and the sides. Especially it is designed as an extensible and modular method which enables to support various additional postures in the future. This system is configured as multiple clients and the Unity3D server. The client processes Kinect skeleton data and send to the server. The server performs the multiple-Kinect calibration process and then applies the pose estimation algorithm based on the Kinect-based posture recognition model using feature extractions and the weighted averaging of feature values for different Kinects. This paper presents the design and implementation of the human exercise pose estimation system using multiple Kinects and also describes how to build and execute an interactive Unity3D exergame.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.39 no.4
• /
• pp.361-367
• /
• 2015

This paper presents the design concept of an intrinsic continuum robot for safe man-machine interface and characteristic behaviors of its end-effector based on real experiments. Since pneumatic artificial muscles having similar antagonistic actuation to human muscles are used for main backbones of the proposed robot as well as in the role of the actuating devices, variable stiffness of robotic joints can be available in the actual environment. In order to solve the inherent shortcoming of an intrinsic continuum robot due to bending motion of the backbone materials, a Kalman filter scheme based on a triaxial accelerometer and a triaxial gyroscope was proposed to conduct an attitude estimation of the end-effector of the robot. The experimental results verified that the proposed method was effective in estimating the attitude of the end-effector of the intrinsic continuum robot.