Sehwan Chun’s research while affiliated with Soongsil University and other places

This study proposes the use of a CNN model to predict the resistance of conductive fabrics by utilizing the brightness information from their images, aiming to address the limitations of traditional contact-based measurement methods and explore the feasibility of non-contact resistance measurement. Conductive fabrics were produced using environmentally friendly cellulose fiber as a base material, with a dip-coating and padding process involving water-based single-walled carbon nanotube (SWCNT). After scanning the produced conductive fabrics and meticulously preprocessing the images, a dataset for CNN training was constructed, comprising label values corresponding to the sheet resistance of each image. ANOVA analysis confirmed a statistically significant relationship (p-value = 8.04145e^-18) between the brightness of conductive fabric images and their sheet resistance. By leveraging the relationship between the brightness of fabric images and sheet resistance, training of the CNN model yielded an RMSE of 0.0558 and an R-squared value of 0.9557, validating the effectiveness of the designed CNN model for image-based resistance prediction. This research is expected to contribute to the development of future real-time monitoring and control systems, providing a crucial foundation for the advancement of data-driven measurement and control systems based on computer vision and machine learning techniques. Furthermore, it is anticipated to unveil new possibilities for various applications of conductive fabrics.

Resistance training is becoming increasingly important and widespread. Decomposition of the muscle loads applied is important for injury prevention and determining the load on the targeted muscles. In this study, a flexible textile PET (polyethylene terephthalate)/SP(Spandex) SWCNT (Single-walled carbon nanotube) stretch sensor was fabricated and attached at four locations: the elbow, brachioradialis/flexor carpi radialis, biceps brachii, and triceps brachii. The stretch sensors attached to the elbow can measure the angle of elbow flexion without an IMU 9-axis sensor using quadratic fitting. A Multi-Layer Perceptron (MLP) was used to decompose the muscle volume expansions of the 3muscle by angle. The model provided a good fit for all three muscles, with R-squared values ranging from Test set 0.98725 to 0.99815. Through one input and three ouput fitting, the muscle volume expansion quantities during the bicep barbell curl were decomposed and compared with data. The results showed that the brachioradialis/flexor carpi radialis muscle maintained 13% of the arm muscle volume up to 60°, then increased to 44% at 100°. The biceps brachii muscle steadily increased to 70% from 0° up to 60°, and then maintained 40% at 100° due to the volume increase of other muscles. The triceps brachii muscle maintained 9% of the arm muscle volume up to 90°, then increased to 20% at 100°. This study shows that muscle volume expansion can be easily measured with a non-body contact wearable device, unlike many existing contact methods for measuring muscle activity like EMG (electro-myography), etc. This study provides a novel approach for easily measuring muscle volume expansion and decomposition in wearable devices, which can indirectly indicate injury prevention and muscle loading in target areas through balance optimization among local muscles.

In this study, the authors proposed a method to fabricate a resistive stretch textile sensor from polyester spandex (PET/SP) fabric and commercial single-walled carbon nanotube (SWCNT). In addition, we designed and trained a one-dimension convolutional neural network to classify four resistance workouts, which employed data acquired from the proposed sensor as the input. To figure out the most appropriate PET/SP sample for the deep learning application, we investigated morphologies and characterization of three samples in distinct conditions of the coating process. Data acquired from the proposed sensor illustrated the significant difference between activated and non-activated muscle groups in each specific exercise. With the PET/SP sample which met the requirements of the application, after 100 epochs, the deep learning model achieved 97.2% training accuracy and 90% test accuracy. This study demonstrates that the SWCNT-coated PET/SP stretch textile sensor can be utilized effectively to track the activity of forearm muscles during resistance training. Other than that, the proposed 1D-CNN, with the advantage of training time and computational cost, is able to classify time series data with high performance and thus can be applied widely in various deep learning applications, especially in the healthcare and sports industries.

Wearables have been applied in the field of fitness in recent years to monitor human muscles by recording electromyographic (EMG) signals. Understanding muscle activation during exercise routines allows strength athletes to achieve the best results. Hydrogels, which are widely used as wet electrodes in the fitness field, are not an option for wearable devices due to their characteristics of being disposable and skin-adhesion. Therefore, a lot of research has been conducted on the development of dry electrodes that can replace hydrogels. In this study, to make it wearable, neoprene was impregnated with high-purity SWCNTs to develop a dry electrode with less noise than hydrogel. Due to the impact of COVID-19, the demand for workouts to improve muscle strength, such as home gyms and personal trainers (PT), has increased. Although there are many studies related to aerobic exercise, there is a lack of wearable devices that can assist in improving muscle strength. This pilot study proposed the development of a wearable device in the form of an arm sleeve that can monitor muscle activity by recording EMG signals of the arm using nine textile-based sensors. In addition, some machine learning models were used to classify three arm target movements such as wrist curl, biceps curl, and dumbbell kickback from the EMG signals recorded by fiber-based sensors. The results obtained show that the EMG signal recorded by the proposed electrode contains less noise compared to that collected by the wet electrode. This was also evidenced by the high accuracy of the classification model used to classify the three arms workouts. This work classification device is an essential step towards wearable devices that can replace next-generation PT.