This study presents the design, optimization, and evaluation of an optical fiber-based force myography (FMG) sensor for hand gesture and grasping force estimation, with potential haptic applications for realistic human-robot interaction (HRI). The sensor, with a diameter of 10 mm and a thickness of 2 mm, is designed to be flexible and easily integrated into clothing or sleeves. Utilizing the principle of light loss due to touch in bent optical fibers, the sensor demonstrates high sensitivity and stability. Through simulation, the sensor design parameters, such as the bending radius and the number of weavings, were optimized to achieve a high signal-to-noise ratio and sensitivity. The fabricated sensor showed excellent performance in various experiments, including high sensitivity, repeatability, and accurate force estimation. A wearable system with four embedded sensors achieved 91% hand gesture recognition accuracy using long short-term memory neural network and grasping force estimation exhibited high linearity across different ball grip types. These findings demonstrate the sensor practicality, with future research focusing on improving durability and exploring broader applications in haptics and HRI.