Conventional technologies to monitor torque feedback and angle in exoskeleton actuators are bulky and sensitive to misalignments, and do not allow for multiplexed operation. Fiber Bragg grating (FBG)-based sensors are a robust sensing approach that are desirable for multi-parametric monitoring. Temperature, strain, torque, and angle are widely studied in human-robot interaction. In order to acquire the torque and angle of deflection in the torsional spring of a series elastic actuator, an experimental setup with the spring and an array of three FBGs is submitted to repeated torques and angles. This paper presents the characterization and validation of the FBG-based sensor for measuring by torque and angle variations. Temperature cross-sensitivity is derived by the use of a non-strain FBG. The developed sensor presented high linearity and small error for torque and angle measurements.