• Journal of the Korean Society for Precision Engineering
• /
• v.28 no.6
• /
• pp.670-677
• /
• 2011

This study is to estimate the joint torques without torque sensor using the EMG (Electromyogram) signal of agonist/antagonist muscle with Neural Network Back Propagation Algorithm during the elbow motion. Command Signal can be guessed by EMG signal. But it cannot calculate the joint torque. There are many kinds of field utilizing Back Propagation Learning Method. It is generally used as a virtual sensor estimated physical information in the system functioning through the sensor. In this study applied the algorithm to obtain the virtual senor values estimated joint torque. During various elbow movement (Biceps isometric contraction, Biceps/Triceps Concentric Contraction (isotonic), Biceps/Triceps Concentric Contraction/Eccentric Contraction (isokinetic)), exact joint torque was measured by KINCOM equipment. It is input to the (BP)algorithm with EMG signal simultaneously and have trained in a variety of situations. As a result, Only using the EMG sensor, this study distinguished a variety of elbow motion and verified a virtual torque value which is approximately(about 90%) the same as joint torque measured by KINCOM equipment.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.10
• /
• pp.5130-5143
• /
• 2019

Football (soccer) is one of the most popular sports in the world. A huge number of people watch live football matches by TV or Internet. A football match takes 90 minutes, but viewers may only want to watch a few highlights to save their time. As far as we know, there is no such a product that can be put into use to achieve intelligent highlight extraction from live football matches. In this paper, we propose an intelligent editing system for live football matches. Our system can automatically extract a series of highlights, such as goal, shoot, corner kick, red yellow card and the appearance of star players, from the live stream of a football match. Our system has been integrated into live streaming platforms during the 2018 FIFA World Cup and performed fairly well.

• Journal of Korean Medical classics
• /
• v.31 no.2
• /
• pp.1-16
• /
• 2018

Objective : The college of Korean medicine uses a variety of Korean medical classics. Thus, in order to overcome the natural difference of the details of their curriculum, this paper aims to study the usage of Flipped Learning as a way of standardizing the curriculum that teach Korean medical classics. Method : One effective teaching and learning methodology, which is called for by the changing educational paradigm, is Flipped Learning. To introduce this learning method, which is actively applied to different curriculum, the paper revises the goal of learning objectives and introduces a teaching model of Flipped Learning in order to suggest the standardization of Korean medical educations through the re-design of the curriculum for Korean medical classics. Result : The professors of the Korean medical classics must work together to use the revised learning objectives and teaching model and create a set of lectures to serve as a basis of educational standardization. Conclusion : The standardization of the education of Korean medical classics through the Flipped Learning method could pre-emptively deal with the Korean medical doctor's capacity model that is in development now.

• ETRI Journal
• /
• v.41 no.5
• /
• pp.560-573
• /
• 2019

Two supervised learning algorithms, a basic neural network and a long short-term memory recurrent neural network, are applied to traffic including DDoS attacks. The joint effects of preprocessing methods and hyperparameters for machine learning on performance are investigated. Values representing attack characteristics are extracted from datasets and preprocessed by two methods. Binary classification and two optimizers are used. Some hyperparameters are obtained exhaustively for fast and accurate detection, while others are fixed with constants to account for performance and data characteristics. An experiment is performed via TensorFlow on three traffic datasets. Three scenarios are considered to investigate the effects of learning former traffic on sequential traffic analysis and the effects of learning one dataset on application to another dataset, and determine whether the algorithms can be used for recent attack traffic. Experimental results show that the used preprocessing methods, neural network architectures and hyperparameters, and the optimizers are appropriate for DDoS attack detection. The obtained results provide a criterion for the detection accuracy of attacks.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.3
• /
• pp.861-880
• /
• 2023

Semantic segmentation of road scene is the key technology of autonomous driving, and the improvement of convolutional neural network architecture promotes the improvement of model segmentation performance. The existing convolutional neural network has the simplification of learning knowledge and the complexity of the model. To address this issue, we proposed a road scene semantic segmentation algorithm based on multi-task collaborative learning. Firstly, a depthwise separable convolution atrous spatial pyramid pooling is proposed to reduce model complexity. Secondly, a collaborative learning framework is proposed involved with saliency detection, and the joint loss function is defined using homoscedastic uncertainty to meet the new learning model. Experiments are conducted on the road and nature scenes datasets. The proposed method achieves 70.94% and 64.90% mIoU on Cityscapes and PASCAL VOC 2012 datasets, respectively. Qualitatively, Compared to methods with excellent performance, the method proposed in this paper has significant advantages in the segmentation of fine targets and boundaries.

• Journal of Institute of Control, Robotics and Systems
• /
• v.10 no.5
• /
• pp.395-401
• /
• 2004

We present an AILC(Adaptive Iterative Learning Control) scheme and a sufficient condition for system parameter identification for uncertain robotic systems that perform the same tasks repetitively. It is guaranteed that the joint velocity and position asymptotically converge to the reference joint velocity and position, respectively. In addition, it is proved that a sufficient condition for parameter identification is the PE(Persistent Excitation) condition on the regressor matrix evaluated at the reference trajectory during the operation period. Since the regressor matrix on the reference trajectory can be easily computed prior to the real robot operation, the proposed algorithm provides a useful method to verify whether the parameter error converges to zero or not.

• 제어로봇시스템학회:학술대회논문집
• /
• 1990.10b
• /
• pp.1173-1177
• /
• 1990

Cerebellar Model Linear Associator Net(CMLAN), a kind of neuro-net based adaptive control function generator, was applied to the problem of direct inverse calibration of three and six d.o.f. POMA 560 robot. Since CMLAN autonomously maps and generalizes a desired system function via learning on the sampled input/output pair nodes, CMLAN allows no knowledge in system modeling and other error sources. The CMLAN based direct inverse calibration avoids the complex procedure of identifying various system parameters such as geometric(kinematic) or nongeometric(dynamic) ones and generates the corresponding desired compensated joint commands directly to each joint for given target commands in the world coordinate. The generated net outputs automatically handles the effect of unknown system parameters and dynamic error sources. On-line sequential learning on the prespecified sampled nodes requires only the measurement of the corresponding tool tip locations for three d.o.f. manipulator but location and orientation for six d.o.f. manipulator. The proposed calibration procedure can be applied to any robot.

• Journal of the Korean Society of Industry Convergence
• /
• v.18 no.4
• /
• pp.191-196
• /
• 2015

In this paper, we present a path-smoothing algorithm for a robot arm manipulator that finds the path using a joint space-based rapidly-exploring random tree. Unlike other smoothing algorithms which require complex mathematical computation, the proposed path-smoothing algorithm is done using a Gaussian process. To find the optimal hyperparameters of the Gaussian process, we use differential evolution hybridized with opposition-based learning. The simulation result indicates that the Gaussian process whose hyperparameters were optimized by hybrid differential evolution successfully smoothed the path generated by the joint space-based rapidly-exploring random tree.

• Journal of the Korean Society for Precision Engineering
• /
• v.12 no.8
• /
• pp.92-98
• /
• 1995

The position control of flexible joint manipulator is investigated by applying the self-organizing fuzzy logic controller (SOC) proposed by Procyk and Mamdani. The SOC is a heuristic rule-based controller and a further extension of an ordinary fuzzy controller, which has a hierachy structrue which consists of an algorithm being identical to a fuzzy controller at the lower ollp and a learning algorithm accomodating the performance evalution and rule modification function at the upper ollp. This form of control can be used in those complex systems which have been too difficult to control or which in the past have had to rely on the experience of a human operator. Even though the significant dynamic coupling of the motors and links on the flexible joint manipulator, the performance of command-following is good by applying the proposed SOC.

• The Journal of Korean Physical Therapy
• /
• v.22 no.6
• /
• pp.43-51
• /
• 2010

Purpose: This study examined the influence of visuomotor congruency on learning a relative phase relationship between a single joint movement and an external signal. Methods: Participants (N=5) were required to rhythmically coordinate elbow flexion-extension movements with a continuous sinusoidal wave (0.375 Hz) at a $90^{\circ}$ relative phase relationship. The congruent group was provided online feedback in which the elbow angle decreased (corresponding to elbow flexion) as the angle trajectory was movingup, and vice versa. The incongruent group was provided online feedback in which the elbow angle decreased as the angle trajectory was moving down, and vice versa. There were two practice sessions (day 1 and 2) and each session consisted of 6 trials per block (5 blocks per session). Retention tests were performed 24 hours after session 2, and only the external sinusoidal wave was provided. Repeated ANOVAs were used for statistical analysis. Results: During practice, the congruent group was significantly less variable than the incongruent group. Phase variability in the incongruent group did not significantly change across blocks, while variability decreased significantly in the congruent group. In retention, the congruent group produced the required $90^{\circ}$ relative phase pattern with significantly less phase variability than the incongruent group. Conclusions: Congruent visual feedback facilitates learning. Moreover, the deprivation of online feedback does not affect the congruent group but does affect the incongruent group in retention.