This study proposes a gait model that describes the relation between the lower limb muscle forces and the center-of-gravity (CoG) velocity in the forward, vertical, and lateral directions. Based on the CoG velocity and lower limb muscle forces obtained from gait measurements, the unknown parameters in the gait model are identified using a Kalman filter. The CoG velocity is estimated by applying the identified parameters and lower limb muscle forces to the model. After the accuracy of the proposed model is verified, the model is used to examine the relationship between the lower limb muscle groups used to construct the gait model and the gait velocity in three directions. The estimated CoG velocity in the vertical and lateral directions is more accurate than that in the forward direction. The analysis results indicate that the fluctuation of the CoG trajectory in the vertical and lateral directions can be accurately expressed using the 10 muscle forces for each leg used in the model. The obtained results can be used to quantitatively evaluate CoG sway. The proposed model can be applied in medical care and sports medicine.