RESUMEN

Taekwondo has evolved from a traditional martial art into an official Olympic sport. This study introduces a novel action recognition model tailored for Taekwondo unit actions, utilizing joint-motion data acquired via wearable inertial measurement unit (IMU) sensors. The utilization of IMU sensor-measured motion data facilitates the capture of the intricate and rapid movements characteristic of Taekwondo techniques. The model, underpinned by a conventional convolutional neural network (CNN)-based image classification framework, synthesizes action images to represent individual Taekwondo unit actions. These action images are generated by mapping joint-motion profiles onto the RGB color space, thus encapsulating the motion dynamics of a single unit action within a solitary image. To further refine the representation of rapid movements within these images, a time-warping technique was applied, adjusting motion profiles in relation to the velocity of the action. The effectiveness of the proposed model was assessed using a dataset compiled from 40 Taekwondo experts, yielding remarkable outcomes: an accuracy of 0.998, a precision of 0.983, a recall of 0.982, and an F1 score of 0.982. These results underscore this time-warping technique's contribution to enhancing feature representation, as well as the proposed method's scalability and effectiveness in recognizing Taekwondo unit actions.

RESUMEN

Issues of fairness and consistency in Taekwondo poomsae evaluation have often occurred due to the lack of an objective evaluation method. This study proposes a three-dimensional (3D) convolutional neural network-based action recognition model for an objective evaluation of Taekwondo poomsae. The model exhibits robust recognition performance regardless of variations in the viewpoints by reducing the discrepancy between the training and test images. It uses 3D skeletons of poomsae unit actions collected using a full-body motion-capture suit to generate synthesized two-dimensional (2D) skeletons from desired viewpoints. The 2D skeletons obtained from diverse viewpoints form the training dataset, on which the model is trained to ensure consistent recognition performance regardless of the viewpoint. The performance of the model was evaluated against various test datasets, including projected 2D skeletons and RGB images captured from diverse viewpoints. Comparison of the performance of the proposed model with those of previously reported action recognition models demonstrated the superiority of the proposed model, underscoring its effectiveness in recognizing and classifying Taekwondo poomsae actions.

RESUMEN

Defect inspection is important to ensure consistent quality and efficiency in industrial manufacturing. Recently, machine vision systems integrating artificial intelligence (AI)-based inspection algorithms have exhibited promising performance in various applications, but practically, they often suffer from data imbalance. This paper proposes a defect inspection method using a one-class classification (OCC) model to deal with imbalanced datasets. A two-stream network architecture consisting of global and local feature extractor networks is presented, which can alleviate the representation collapse problem of OCC. By combining an object-oriented invariant feature vector with a training-data-oriented local feature vector, the proposed two-stream network model prevents the decision boundary from collapsing to the training dataset and obtains an appropriate decision boundary. The performance of the proposed model is demonstrated in the practical application of automotive-airbag bracket-welding defect inspection. The effects of the classification layer and two-stream network architecture on the overall inspection accuracy were clarified by using image samples collected in a controlled laboratory environment and from a production site. The results are compared with those of a previous classification model, demonstrating that the proposed model can improve the accuracy, precision, and F1 score by up to 8.19%, 10.74%, and 4.02%, respectively.

Asunto(s)

RESUMEN

PURPOSE: We compared and analyzed energy consumption and excess post-exercise oxygen consumption (EPOC) following Taekwondo Taegeuk Poomsae performances. METHODS: Forty-two healthy men who could perform Taegeuk Poomsae 1-8 Jangs were enrolled in this study. To reduce the impact of Poomsae, a random cross-design was used. The washout time was set to at least three days. Oxygen consumption (VO2) was monitored after performing each Poomsae until a reference line was resumed. Each Taegeuk Poomsae was performed at a speed of 60 bpm. RESULTS: There was no significant difference in VO2, carbon dioxide excretion, and heart rate after performing the Taegeuk Poomsae once; however, all variables increased significantly in combined results of EPOC metabolism (F<45.646, p<0.001, and ɳ2<0.527). Taegeuk 8 Jang had the highest levels of all the factors. There were noticeable variations in the oxidation of fat and carbohydrates throughout the Taegeuk Poomsae (F<9.250, p<0.001, ɳ2<0.184). Taegeuk 8 Jang demonstrated the greatest rate of carbohydrate oxidation, and 4-8 Jangs demonstrated much greater rates of fatty acid oxidation. Compared to 1 Jang, the energy consumption showed significant differences in all the variables and peaked in Taegeuk 8 Jang. CONCLUSION: The energy consumption during the Poomsae performances was the same. When the EPOC metabolism was coupled, it was evident that more energy was substantially used in each chapter of Poomsae. Consequently, it was determined that when performing Poomsae, not only should energy metabolism during exercise be taken into account but also EPOC metabolism, which can increase by 10-fold.

RESUMEN

Heart rate variability (HRV) is an effective tool for objectively evaluating physiological stress indices in psychological states. This study aimed to develop multiple linear regression equations to predict HRV variables using physical characteristics, body composition, and heart rate (HR) variables (eg, sex, age, height, weight, body mass index, fat-free mass, percent body fat, resting HR, maximal HR, and HR reserve) in Korean adults. Six hundred eighty adults (male, n = 236, female, n = 444) participated in this study. HRV variable estimation multiple linear regression equations were developed using a stepwise technique. The regression equation's coefficient of determination for time-domain variables was significantly high (SDNN = adjusted R2: 73.6%, P < .001; RMSSD = adjusted R2: 84.0%, P < .001; NN50 = adjusted R2: 98.0%, P < .001; pNN50 = adjusted R2: 99.5%, P < .001). The coefficient of determination of the regression equation for the frequency-domain variables was high without VLF (TP = adjusted R2: 75.0%, P < .001; LF = adjusted R2: 77.6%, P < .001; VLF = adjusted R2: 30.1%, P < .001; HF = adjusted R2: 71.3%, P < .001). Healthcare professionals, researchers, and the general public can quickly evaluate their psychological conditions using the HRV variables prediction equation.

Asunto(s)

RESUMEN

Measuring functional fitness (FF) to track the decline in physical abilities is important in order to maintain a healthy life in old age. This paper aims to develop an estimation model of FF variables, which represents strength, flexibility, and aerobic endurance, using easy-to-measure physical parameters for Korean older adults aged over 65 years old. The estimation models were developed using various machine learning techniques and were trained with the National Fitness Award datasets from 2015 to 2019 compiled by the Korea Sports Promotion Foundation. The machine-learning-based nonlinear regression models were employed to improve the performance of the previous linear regression models. To derive the optimal estimation model that showed the best estimation accuracy, we developed five different machine-learning-based estimation models and compares the estimation accuracy not only among the machine learning models, but also with the previous linear regression model. The coefficient of determination of the FF variables was used to compare the performance of each model; the mean absolute percentage error (MAPE) and standard error of estimation (SEE) were used to evaluate the model performance. The deep neural network (DNN) model presented the best performance among the regression models for the estimation of all of the FF variables. The coefficient of determination in the HGS test was 0.784, while those of the others were less than 0.5 meaning that the HGS of older adults can be reliably estimated using easy-to-measure independent variables.

Asunto(s)

RESUMEN

The main advantage of measuring functional fitness (FF) in older adults is that individual tests can estimate and track the rate of decline with age. This study aimed to develop a multiple linear regression model for predicting FF variables using easy-to-measure independent variables (e.g., sex, age, body mass index, and percent body fat) in Korean older adults. National Fitness Award datasets from the Republic of Korea were used in this analysis. The participants were aged ≥65 years and included 61,465 older men and 117,395 older women. The FF variables included the hand grip strength, lower body strength (30-s chair stand), lower body flexibility (chair sit-and-reach), coordination (figure of 8 walk), agility/dynamic balance (timed up-and-go), and aerobic endurance (2-min step test). An estimation multiple linear regression model was developed using a stepwise technique. In the regression model, the coefficient of determination in the hand grip strength test (adjusted R2 = 0.773, p < 0.001) was significantly high. However, the coefficient of determination in the 30-s chair stand (adjusted R2 = 0.296, p < 0.001), chair sit-and-reach (adjusted R2 = 0.435, p < 0.001), figure of 8 walk (adjusted R2 = 0.390, p < 0.001), timed up-and-go (adjusted R2 = 0.384, p < 0.001), and 2-min step tests (adjusted R2 = 0.196, p < 0.001) was significantly low to moderate. Our findings suggest that easy-to-measure independent variables can predict the hand grip strength in older adults. In future studies, explanatory power will be further improved if multiple linear regression analysis, including the physical activity level and nutritional status of older adults, is performed to predict the FF variables.

RESUMEN

Target-following mobile robots have gained attention in various industrial applications. This study proposes an ultra-wideband-based target localization method that provides highly accurate and robust target tracking performance for a following robot. Based on the least square approximation framework, the proposed method improves localization accuracy by compensating localization bias and high-frequency deviations component by component. Initial calibration method is proposed to measure the device-dependent localization bias, which enables a compensation of the bias error not only at the calibration points, but also at the any other points. An iterative complementary filter, which recursively produces optimal estimation for each timeframe as a weighted sum of previous and current estimation depending on the reliability of each estimation, is proposed to reduce the deviation of the localization error. The performance of the proposed method is validated using simulations and experiments. Both the magnitude and deviation of the localization error were significantly improved by up to 77 and 51%, respectively, compared with the previous method.

RESUMEN

Estimation of health-related physical fitness (HRPF) levels of individuals is indispensable for providing personalized training programs in smart fitness services. In this study, we propose an artificial neural network (ANN)-based estimation model to predict HRPF levels of the general public using simple affordable physical information. The model is designed to use seven inputs of personal physical information, including age, gender, height, weight, percent body fat, waist circumference, and body mass index (BMI), to estimate levels of muscle strength, flexibility, maximum rate of oxygen consumption (VO2max), and muscular endurance. HRPF data (197,719 sets) gathered from the National Fitness Award dataset are used for training (70%) and validation (30%) of the model. In-depth analysis of the model's estimation accuracy is conducted to derive optimal estimation accuracy. This included input/output correlation, hidden layer structures, data standardization, and outlier removals. The performance of the model is evaluated by comparing the estimation accuracy with that of a multiple linear regression (MLR) model. The results demonstrate that the proposed model achieved up to 10.06% and 30.53% improvement in terms of R2 and SEE, respectively, compared to the MLR model and provides reliable estimation of HRPF levels acceptable to smart fitness applications.

Asunto(s)

RESUMEN

OBJECTIVE: To evaluate a novel multi-channel functional electrical stimulation (FES) rehabilitation method based on the evaluation of patient-specific walking dysfunction. METHODS: This study investigated a novel multi-channel FES-based rehabilitation method that analysed the patient's muscle synergy and walking posture. A patient-specific FES profile was produced in the pre-evaluation stage by comparing the muscle synergy and walking posture of the patient with those of healthy control subjects. During the rehabilitation phase, this profile was used to determine an appropriate FES pulse width and amplitude for stimulating the patient's muscles as they walked across a flat surface. RESULTS: Two stroke patients with hemiplegic symptoms participated in a clinical evaluation of the proposed method involving a 4-week course of rehabilitation. An evaluation of the rehabilitation results based on a comparison of the pre- and post-rehabilitation muscle synergy and walking posture revealed that the rehabilitation enhanced the muscle synergy similarity between the patients and healthy control subjects and their quantitative walking performance, as measured by a 10-m walk test and walking speed, by up to 23.38% and 30.00%, respectively. CONCLUSION: These results indicated that the proposed rehabilitation method improved walking ability by improving muscle coordination and adequately supporting weakened muscles in stroke patients.

Asunto(s)

RESUMEN

Continuous health care and the measurement of health-related physical fitness (HRPF) is necessary for prevention against chronic diseases; however, HRPF measurements including laboratory methods may not be practical for large populations owing to constraints such as time, cost, and the requirement for qualified technicians. This study aimed to develop a multiple linear regression model to estimate the HRPF of Korean adults, using easy-to-measure dependent variables, such as gender, age, body mass index, and percent body fat. The National Fitness Award datasets of South Korea were used in this analysis. The participants were aged 19-64 years, including 319,643 male and 147,600 females. HRPF included hand grip strength (HGS), flexibility (sit and reach), muscular endurance (sit-ups), and cardiorespiratory fitness (estimated VO2max ). An estimation multiple linear regression model was developed using the stepwise technique. The outlier data in the multiple regression model was identified and removed when the absolute value of the studentized residual was ≥2. In the regression model, the coefficient of determination for HGS (adjusted R 2: 0.870, P < 0.001), muscular endurance (adjusted R 2: 0.751, P < 0.001), and cardiorespiratory fitness (adjusted R 2: 0.885, P < 0.001) were significantly high. However, the coefficient of determination for flexibility was low (adjusted R 2: 0.298, P < 0.001). Our findings suggest that easy-to-measure dependent variables can predict HGS, muscular endurance, and cardiorespiratory fitness in adults. The prediction equation will allow coaches, athletes, healthcare professionals, researchers, and the general public to better estimate the expected HRPF.

RESUMEN

A haptic interface based on electrical muscle stimulation (EMS) has huge potential in terms of usability and applicability compared with conventional haptic interfaces. This study analyzed the force response characteristics of forearm extensor muscles for EMS-based haptic rendering. We introduced a simplified mathematical model of the force response, which has been developed in the field of rehabilitation, and experimentally validated its feasibility for haptic applications. Two important features of the force response, namely the peak force and response time, with respect to the frequency and amplitude of the electrical stimulation were identified by investigating the experimental force response of the forearm extensor muscles. An exponential function was proposed to estimate the peak force with respect to the frequency and amplitude, and it was verified by comparing with the measured peak force. The response time characteristics were also examined with respect to the frequency and amplitude. A frequency-dependent tendency, i.e., an increase in response time with increasing frequency, was observed, whereas there was no correlation with the amplitude. The analysis of the force response characteristics with the application of the proposed force response model may help enhance the fidelity of EMS-based haptic rendering.

Asunto(s)

RESUMEN

In taekwondo, poomsae (i.e., form) competitions have no quantitative scoring standards, unlike gyeorugi (i.e., full-contact sparring) in the Olympics. Consequently, there are diverse fairness issues regarding poomsae evaluation, and the demand for quantitative evaluation tools is increasing. Action recognition is a promising approach, but the extreme and rapid actions of taekwondo complicate its application. This study established the Taekwondo Unit technique Human Action Dataset (TUHAD), which consists of multimodal image sequences of poomsae actions. TUHAD contains 1936 action samples of eight unit techniques performed by 10 experts and captured by two camera views. A key frame-based convolutional neural network architecture was developed for taekwondo action recognition, and its accuracy was validated for various input configurations. A correlation analysis of the input configuration and accuracy demonstrated that the proposed model achieved a recognition accuracy of up to 95.833% (lowest accuracy of 74.49%). This study contributes to the research and development of taekwondo action recognition.

Asunto(s)

RESUMEN

The vehicular ad hoc network (VANET) has been identified as one of the most promising technologies for managing future intelligent transportation systems. This paper proposes a distributed transmission power adjustment algorithm for communication congestion control and awareness enhancement to address communication congestion problems that can arise in VANETs. The objective of the proposed algorithm is to provide maximum awareness of surrounding vehicles' status while maintaining a communications channel load below the allowed threshold. The proposed algorithm accomplishes this by adjusting the transmission range of each vehicle in the network progressively and gradually, while monitoring the communications channel load of each vehicle. By changing the transmission range of a vehicle little by little according to the communications channel load of its neighboring vehicles, the algorithm finds the optimal transmission range that provides maximum awareness without resulting in communications congestion. In addition, the proposed algorithm appropriately controls the channel load in a fair manner without sacrificing awareness of specific vehicles in the congested situation. This allows nearby vehicles to obtain more peripheral information to help them stay away from potential hazards and maintain safety. The proposed algorithm was implemented in a simulation environment, and its performance was validated in various traffic scenarios. The simulation results show that the proposed algorithm can deal with communication congestion by controlling the transmission power fairly to a target threshold in various traffic situations.

Asunto(s)

RESUMEN

This paper proposes a markerless tracking method with adaptive pose estimation for augmenting 3D organ models on top of the endoscopic image for Endoscopic Retrograde Cholangiopancreatography (ERCP). While many applications of augmented reality (AR) to surgeries need special markers to track the camera's position and orientation in the live video, our method employs the feature detection techniques to track the endoscopic camera. One of the most difficult problems when applying feature-based method to AR for ERCP is the lack of texture & highly specular reflection surface of duodenum in the endoscopic images, which does not provide a stable number of keypoints to track in the endoscopic video sequence. By introducing an adaptive weight function in the combination of reference-current frame tracking with previous-current frame tracking, we enhance the tracking performance remarkably. The proposed method is evaluated using an endoscopic video of a real ERCP and 3D duodenum model reconstructed from CT data of the patient. The result shows real-time performance and robustness of the method.

Asunto(s)

RESUMEN

This paper proposes a novel simulation framework for the real-time deformation of the colon and endoscope using a skeleton-driven deformation method. Cylindrical lattices and a centerline are employed as the skeletons, and a mass-spring model is applied to the skeletons for the mechanics-based simulation. The centerline-based collision detection and resolution algorithm is applied to simulate the interaction between the colon and endoscope. The proposed simulation framework is integrated with a colonoscopy simulation. Simulation results show that the proposed method allows real-time simulation (28 Hz) using the colon model composed of up to 241,440 meshes.

Asunto(s)

RESUMEN

BACKGROUND: Colonoscopy simulation has been increasingly applied as a method of training, which can supplement the traditional patient-based training in recent years. However, the current level of realism of the simulation is insufficient. One of the main difficulties degrading the realism is real-time simulation of colon deformation involving multi-contact interaction with the endoscope. METHODS: This paper proposes a novel simulation framework for real-time deformation of the colon and endoscope, using a skeleton-driven deformation method. Cylindrical lattices and a centre-line are employed as the skeletons, and a mass-spring model is applied to the skeletons for the mechanics-based simulation. The centre-line-based collision detection and resolution algorithm is proposed to simulate the interaction between the colon and endoscope. A haptic rendering algorithm using the energy method is proposed to produce feedback force, based on physical interaction between the colon and endoscope. RESULTS: The proposed simulation framework has been implemented and evaluated in colonoscopy simulation. The simulation results show that the proposed method allows real-time simulation (28 Hz) using a colon model composed of up to 241,440 meshes. CONCLUSIONS: The proposed method allows real-time simulation of colon and endoscope deformation while maintaining a visually plausible result and realistic haptic sensation.

Asunto(s)

RESUMEN

This article describes a series of contributions in the field of real-time simulation of soft tissue biomechanics. These contributions address various requirements for interactive simulation of complex surgical procedures. In particular, this article presents results in the areas of soft tissue deformation, contact modelling, simulation of cutting, and haptic rendering, which are all relevant to a variety of medical interventions. The contributions described in this article share a common underlying model of deformation and rely on GPU implementations to significantly improve computation times. This consistency in the modelling technique and computational approach ensures coherent results as well as efficient, robust and flexible solutions.

Asunto(s)

RESUMEN

The goal of this study is to validate the KAIST-Ewha Colonoscopy Simulation II as a training tool by examining the sectional learning curve of the trainees' performance on the simulation. Nine subjects including three fellows and six residents in the internal medicine participated in this study. All the subjects practiced the colonoscopy on the simulation until their performance surpasses the criteria preset by colonoscopy experts. Performance of the subjects during all the trials was measured in terms of eight performance indices and analyzed according to the colon segments. The results show that the trainees' skills significantly improved through training on the KAIST-Ewha Colonoscopy Simulation II. Particularly, most of the improvement appeared in the sigmoid and descending colon. On the other hand, there was little improvement in the ascending colon.

Asunto(s)

RESUMEN

This paper presents a surface-data-based haptic rendering method for simulation of surgery of closed reduction and internal fixation (CRIF). Volumetric data is often employed in the simulation of bone surgery because the volume rendering can easily handle information such as density and rigidity of each voxel. However, it is difficult to implement real-time graphics and haptic rendering because of the large computational workload. Therefore, we propose a surface-data-based haptic rendering method for real-time rendering. Mechanical properties and graphics of the inner part of the bone should be modeled in addition to the surface data to simulate drilling into the bone. An algorithm is developed to construct the surface of the drilled hole. This method allows the user of the simulation to feel the varying forces according to the drilled depth.

Asunto(s)