Velocity and position are mostly used signals in structural control and monitoring applications. While in real applications, acceleration measurements are easily accessible using accelerometers. Velocity and position are then estimated by integrating the measured acceleration. This paper use two techniques, baseline correction and low-pass filtering, to solve the problems associated with the numerical integration of real-time acceleration measurements. Although these two techniques are well-known, previous approaches have not combined them for online velocity and position estimation. Using both techniques allows improving integration accuracy. Experimental results on a shake table illustrate the effectiveness of the proposed method.