The purpose of this study was to clarify the stiffness of the anterior tibial muscle in the stance and swing phases during walking. Electrical stimulation was applied to the common peroneal nerve once every two steps at heel strike, and in the stance and swing phases. Mechanomyograms (MMGs) of the anterior tibial muscle were measured with an acceleration sensor. The measured MMGs were divided into stimulated and non-stimulated MMGs, and each set of MMGs was synchronously averaged. A Kalman filter was constructed by approximating the non-stimulated MMG with an autoregressive model. The stimulated MMG was smoothed with the Kalman filter, and then the walking acceleration was obtained. Evoked MMG was extracted by subtracting the walking acceleration from the stimulated MMG. The transfer function from electrical stimulation to evoked MMG was identified using a singular value decomposition method. The natural frequency as an index of muscle stiffness was calculated from the poles of the transfer function. The natural frequency of the evoked MMG during rest was also calculated. The evoked MMGs at heel strike, stance, and swing phases and during rest were approximated well with the sixth-, sixth-, eighth-, and sixth-order models, respectively. The higher order model in the swing phase may reflect the complex vibrations involving the foot and tendons. The natural frequency was highest at heel strike, followed by the stance phase, swing phase, and at rest. Stiffness of the anterior tibial muscle at the stance and swing phases in walking was clarified by the proposed method.