The human musculoskeletal system offers a flexible and stable walking ability by constantly changing the stiffness of the ankle joint during walking. In order to imitate that movement of the ankle, the stiffness estimation of the ankle during movement is important. In this study, EMG-based stiffness estimation of the ankle is presented and it has been shown that the robot can mimic the stiffness value of the user in real time by providing low interaction torques between VS-AnkleExo-user with the applied force feedback impedance control algorithm. To describe the behavior of the ankle joint, a musculoskeletal model approach consisting of a joint driven by two muscles was used. Mykin muscle model was used to create muscle forces that will provide plantar-flexion and dorsal flexion movements of the ankle. Then, the parameters in the Mykin model were determined by using the EMG data obtained by different feature extraction methods and the measured torque data. Since the estimation of these parameters differs according to signal processing methods, it was decided that the Slope Sign Change signal processing method is the most suitable one with the verification experiment. The biomechanical parameters found with this method were replaced in the equations obtained with the help of Mykin muscle model and the stiffness estimation of the ankle joint was performed. Finally, within the scope of the study, the estimated stiffness value of the ankle joint was sent to the stiffness adjustment mechanism of VS-AnkleExo in real time and an impedance control algorithm with force feedback was implemented on the device to obtain minimum interaction torque between the user and VS-AnkleExo.