• The Journal of Korea Robotics Society
• /
• v.17 no.3
• /
• pp.264-272
• /
• 2022

Human-robot cooperative tasks are increasingly required in our daily life with the development of robotics and artificial intelligence technology. Interactive reinforcement learning strategies suggest that robots learn task by receiving feedback from an experienced human trainer during a training process. However, most of the previous studies on Interactive reinforcement learning have required an extra feedback input device such as a mouse or keyboard in addition to robot itself, and the scenario where a robot can interactively learn a task with human have been also limited to virtual environment. To solve these limitations, this paper studies training strategies of robot that learn table balancing tasks interactively using deep reinforcement learning with human's facial expression feedback. In the proposed system, the robot learns a cooperative table balancing task using Deep Q-Network (DQN), which is a deep reinforcement learning technique, with human facial emotion expression feedback. As a result of the experiment, the proposed system achieved a high optimal policy convergence rate of up to 83.3% in training and successful assumption rate of up to 91.6% in testing, showing improved performance compared to the model without human facial expression feedback.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.4
• /
• pp.619-629
• /
• 2013

This study was conducted to compare the effect improving balance abilities of 3 balance training programs by randomized intervention trial. Study subjects were 29 female students recruited from a university in Gwanju, who allocated randomly to 3 intervention groups, aqua group(modified Halliwick 10 point program, N=10), visual bio-feedback balance training group(N=10), and balance-pad plus trampolin group(N=9). As the results of repeated musure ANOVA, Wilcoxon signed rank test, major research findings were as follows; 1. Regarding SI or BIEMG measured at upright posture on hard platform, bio-feedback group under the condition with opening both eye and aqua group under the condition with closing both eye showed signigicant improvement in balance ability, respectively. 2. Regarding SI or BIEMG measured at upright posture on soft platform, bio-feedback group under the condition with opening both eye and aqua group under the condition with closing both eye showed signigicant improvement in balance ability, respectively. 3. Regarding all 3 sensory ratios, bio-feedback group showed significantly higher increasement in them than other 2 groups. The results suggested that visual bio-feedback training program improve the all 3 sensory functions required for balancing, and aqua training modified from Halliwick 10 point program contribute to improve vestibular function principally.

• Journal of International Academy of Physical Therapy Research
• /
• v.3 no.1
• /
• pp.370-377
• /
• 2012

This study examines the effects of different environments on the application of hemiplegia patients circuit balance training. Group 1 performed circuit balance training without any auditory intervention Group 2 performed training in noiseless environments and Group 3 performed training in white noise environments. First, among lower extremity muscular strength evaluation items, maximum activity time(MAT) was not significantly different(p>.05). Maximum muscle strength(MMS) increased significantly in Group 3(p<.01), there was no significant difference in MMS among the groups. Average muscle strength(AMS) indexes also significantly increased in Group 3(p<.01), there was no significant difference in AMS among the groups. Second, among balancing ability evaluation items, Berg's balance scale(BBS) scores significantly increased in all groups(p<.05), BBS scores were significantly difference among the groups. Based on the results, Group 1, 2 and Group 1, 3 showed significant increases (p<.05). Functional reach test(FRT) values significantly increased in Group 2, 3(p<.05), and there was no significant difference in FRT values among the groups. Timed up and go(TUG) test values significantly decreased in Group 2, 3(p<.05), and there was no significant difference in TUG test values among the groups. Third, among walking speed evaluation items, the time required to walk 10m significantly decreased in all groups(p<.05), and there was no significant difference in the values among the groups. Average walking speeds showed significant increases in Group 1, 3(p<.05), and there was no significant difference in the values among the groups. Based on the results of this study, noise environments should be improved by either considering auditory interventions and noiseless environments, or by ensuring that white noise environments facilitate the enhancement of balancing ability.

• The Journal of Korean Physical Therapy
• /
• v.28 no.3
• /
• pp.189-194
• /
• 2016

Purpose: This study focuses on influence of ankle stabilization training on balance ability and lower limb muscle activation of soccer player with functional ankle instability. Methods: Subjects were grouped into ankle stabilization training group using biofeedback comprised of 15 subjects and general exercise group of 15. The training was conducted for 30 minutes, 3 times a week for 8 weeks in total. All 30 football players conducted plyometric training for 30 minutes before main training. To evaluate balance ability, biorescure was used to measure whole path length and surface area and surface electromyography (EMG) system was used to measure tibialis anterior, tibialis posterior, and soleus to evaluate lower limb muscle activation. Results: The experiment group showed significant difference to the comparison group in regard of whole path length and surface area which represents balancing capability and muscle activation of tibialis anterior, tibialis posterior, and soleus. Conclusion: Therefore, ankle stabilization training using biofeedback is more effective in enhancing balance ability and lower limb muscle activation than general exercise.

• PNF and Movement
• /
• v.15 no.1
• /
• pp.27-33
• /
• 2017

Purpose: This study examined the effects of upper- and lower-limb coordinated exercise with proprioceptive neuromuscular facilitation (PNF) on stroke patients' recovery of their balancing and walking abilities. Methods: This study was conducted with 30 patients aged at least 60 years and diagnosed with stroke. The patients were randomly assigned to either a PNF upper- and lower-limb coordinated exercise group of 15 patients or an aero-step balance exercise group of 15 patients. To test the subjects' balancing and walking abilities, balancing ability tests and 10-m walking speed tests were conducted before and after the interventions. The patients performed their respective exercises for 30 minutes per session, three times per week for four weeks. The PNF exercise group performed six stages of exercise consisting of a combination of PNF patterns such as sprinting, skating, and striking. The six stages (right striking, right skating, right sprinting, left striking, left skating, and left sprinting) were performed continuously with a rest period of 1 min. after training for 4 min. The exercises for the aero-step balance group consisted of balancing in a two-leg standing position, weight shifting in a two-leg standing position, one-leg standing, squat exercises, marching in place, and squatting on an aero step. Results: The PNF exercise group showed significant improvements in their balancing ability evaluations compared to the aero-step balance group and also showed significant improvements in their 10-m walking speed tests. Conclusion: Based on the results of this study, PNF upper- and lower-limb coordinated exercise resulted in clinical improvements of stroke patients' balancing and walking abilities. Therefore, this type of exercise is recommended as a clinical intervention for the recovery of stroke patients' lower-limb function. Future studies should be conducted with longer intervention periods and more subjects to generalize the study results.

• Journal of information and communication convergence engineering
• /
• v.11 no.4
• /
• pp.268-273
• /
• 2013

In this paper, we propose a document classification model using Web documents as a part of the training corpus in order to resolve the imbalance of the training corpus size per category. For the purpose of retrieving the Web documents closely related to each category, the proposed document classification model calculates the matching score between word features and each category, and generates a Web search query by combining the higher-ranked word features and the category title. Then, the proposed document classification model sends each combined query to the open application programming interface of the Web search engine, and receives the snippet results retrieved from the Web search engine. Finally, the proposed document classification model adds these snippet results as Web documents to the training corpus. Experimental results show that the method that considers the balance of the training corpus size per category exhibits better performance in some categories with small training sets.

• Journal of the Korea Convergence Society
• /
• v.9 no.9
• /
• pp.143-148
• /
• 2018

The objective of the study was the effects of balance training using sandals on the balancing and proprioception of patients with chronic ankle instability(CAI). A total of 36 patients with CAI were assigned to sandal exercise(SE), balance training(BT), and control groups(CG). The groups were assessed for static balancing and proprioception before and after the intervention. The SE performed using balance sandals, while BT performed balance training. Intervention comprised a 30minute session. 3 times a week for 8 weeks. The SE showed significant effect in static balance, as compared to the other two groups (p<.001), whereas proprioception showed a significant effect on eversion (p<.001). Therefore, balance training using sandals can be a positive intervention method for prevention of falls in elderly as well as chronic ankle instability patients.

• Industrial Engineering and Management Systems
• /
• v.4 no.1
• /
• pp.36-46
• /
• 2005

In traditional production lines, such as assembly lines, each worker is usually assigned to a fixed task, which is beneficial since it reduces the amount of training needed for workers to master their assigned tasks. However, when workers complete their tasks at different speeds, the slowest worker will determine the overall pace of the production line and limit production. To avoid this problem, the self-balancing production line was introduced. In this type of production line, each worker works dynamically, thus they can maintain balanced production. Previous research analyzing the performance of these lines has ignored the walk-back time associated with dynamic workers. U-shaped production lines have also been analyzed and policies for such lines have been proposed. However, the walk-back time cannot be ignored in practice, and research taking this factor into account is needed to enable balanced production and thus the maximum production rate. In this paper, we propose production policies for a production line with the walk-back time taken into account, and define and analyze the conditions for self-balancing. Furthermore, we have compared the performance of such a line with that of other production lines under the same conditions, and the results show the superiority of this line in certain cases.

• PNF and Movement
• /
• v.21 no.1
• /
• pp.119-128
• /
• 2023

Purpose: Obstacle crossing training is being used to improve the walking ability of stroke patients, but studies on which method is more effective when performing obstacle crossing training with an unaffected limb lead (OCT-ULL) and an affected limb lead (OCT-ALL) are not well known. As such, this study aims to compare the intervention effects of obstacle crossing training using unaffected limb leads (OCT-ULL) and obstacle crossing training using affected limb leads (OCT-ALL). Methods: In total, 25 patients with chronic stroke were studied and assigned randomly to the obstacle crossing training with unaffected limb leads (OCT-ULL) group or the obstacle crossing training with affected limb leads (OCT-ALL) group. A lower extremity strength test, balance and gait test, and fall efficacy test were conducted as preliminary tests, and all patients participated in the intervention for 30 minutes a day, five days a week for four weeks, and the same preliminary tests were conducted post-intervention. Results: Compared with the OCT-ALL group, the OCT-ULL group showed a significant improvement in the strength of the affected hip abductor muscle and in balance and gait, as well as in fall efficacy (p<.05). Conclusion: This study suggested that applying the OCT-ULL training method in the obstacle crossing training of stroke patients is more effective for improving balance and gait functions than OCT-ALL.

• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.2
• /
• pp.161-168
• /
• 2024

As the size of big data models grows, distributed training is emerging as an essential element for large-scale machine learning tasks. In this paper, we propose ParamHub for distributed data training. During the training process, this agent utilizes the provided data to adjust various conditions of the model's parameters, such as the model structure, learning algorithm, hyperparameters, and bias, aiming to minimize the error between the model's predictions and the actual values. Furthermore, it operates autonomously, collecting and updating data in a distributed environment, thereby reducing the burden of load balancing that occurs in a centralized system. And Through communication between agents, resource management and learning processes can be coordinated, enabling efficient management of distributed data and resources. This approach enhances the scalability and stability of distributed machine learning systems while providing flexibility to be applied in various learning environments.