Abstract: To address the challenge of determining walking speed parameters in the joint trajectory planning of lower limb prostheses and to help the prostheses to realize better locomotion modes conversion and dynamic motion, we propose a method of walking speed calculation based on the gait phase and a rolling foot model. Initially, we filter and normalize the angular velocity and angle data from the thigh's sagittal plane and calculate the walking frequency using the gait phase variable. Next, using the lower limb rolling foot model during human walking, we establish a functional relationship among step length, real-time angle, and leg length. We then calculate walking speed by multiplying step length and step frequency over different cycles. Experimental results show that within a speed range of 2.0 to 4.0 km/h, the proposed method yields an average root mean square error of 0.054 m/s in walking speed calculation. It indicates that the proposed method has better static and dynamic walking speed calculation performance, and can provide real-time and reliable walking speed parameters.