With the further development of intelligent robots, bionic touch sensing has become an important means for the robotic dexterous hand to interact with the external environment, making it a hot topic. However, in the existing research on tactile perception of robotic hands, most of the sensing units have complex structures and are difficult to integrate in the hand.For touch sensing of the dexterous robot hand, this work employs optical Fiber Bragg Grating (FBG) as the information transmission and sensing carrier and develops a wearable touch fingerstall. The thumb fingerstall contains two segments: fingertip and finger pulp. The fingertip is composed of a curved surface and a plane surface. The curved surface can fit the thumb shape better, and the plane surface can provide a more stable touch. Specially, we use a cylindrical pin as the knuckle to connect the fingertip and finger pulp. The touch sensing of three contact positions of the fingerstall have been studied:position 1 is the vertical fingertip, position 2 is the 15° fingertip, and position 3 is the 30° finger pulp. Firstly, the fingerstall model was simulated using the finite element method to verify its structural feasibility. Then, the fingerstall was loaded using an electronic tension machine, ranging from 0 N to 10 N with a 1 N increment, and each load was held for 25 s. For each test position, the experiment was repeated three times, and the responses of FBG were recorded. Analysis of the touch sensing experimental data shows that the sensing fingerstall has good sensitivity and linearity within the 0~10 N touch pressure range. The average sensitivities at position 1, position 2, and position 3 are 24.119 6 pm/N, 10.338 3 pm/N, and -1.580 7 pm/N, respectively, with linearity above 99%. Overall, the simulation results are basically consistent with the experimental results, although some errors may arise due to 3D printing and FBG packaging. Therefore, different loading positions and forces can be distinguished by different center wavelength offset and offset direction of FBG.In order to verify the sensing performance of the fingerstall further, a hardness discrimination experiment was carried out primarily focused on perceiving objects of small hardness. And the standard ratio of wavelength to touch depth (k_s) was proposed as a quantitative measure of hardness perception. Position 1 was selected as the experimental position for hardness perception, and four silicone blocks with different hardness (named as 0 HA, 10 HA, 20 HA and 40 HA) were selected as the experimental objects. Considering the time and safety factors, 5 mm was selected as the touching depth, and 60mm/min as the touching speed. The average responses of FBG to 0 HA, 10 HA, 20 HA and 40 HA are 58.57 pm, 242.3 pm, 580.95 pm, 1107.6 pm, respectively. The corresponding k_s values are 11.717 pm/mm, 48.46 pm/mm, 116.19 pm/mm and 221.52 pm/mm, respectively. According to the data results, when the fingerstall touches the 40° silicone block, the k_s is the largest, and the k_s is the smallest at the 0° silicone block. When k_s is larger, the greater the hardness of the object is. The sensing fingerstall has a good repetition, with repeatability errors of 3.82%, 0.97%, 0.51%, and 0.29%.In conclusion, a simple and wearable fingerstall has been designed. The tactile sensitivity of three different position has been studied, and a hardness discrimination experiment has been conducted. These research results can serve as a basis for bionic touch sensing of dexterous robotic hands.