Wearable Fiber Sensor for Hand Gesture Recognition

This project involved the development of a hand gesture classification system using fiber sensors that change resistance based on length. Previous research often relied on attaching sensors to gloves, this approach had several limitations. For instance, gloves could interfere with vision-based systems, making it difficult for cameras or other vision technologies to recognize the hand. Additionally, wearing gloves can be uncomfortable and impractical for extended use.

To address these issues, we proposed a novel approach: storing the sensors on the wrist and extending them to attach to the fingers. This method allows for precise measurement of finger movements, enabling accurate hand gesture recognition without the need for bulky or restrictive equipment. By leveraging this design, we achieved a more seamless and user-friendly solution for capturing and interpreting hand gestures.

The training process consists of two stages: the first stage focuses on understanding the relationship between sensor data and hand joint positions, while the second stage correlates sensor data with specific gestures. To address challenges such as user and sensor variability, we incorporated few-shot learning techniques, allowing the system to adapt quickly to new users and sensors with minimal additional data. The final system, tested and validated using Unity VR scene, demonstrates high accuracy in gesture classification, showcasing the potential of integrating advanced sensor technology with machine learning for intuitive human-computer interaction.

Data Collection