Optimizing the error between the estimated signal and expected signal is the major goal of a filtering algorithm and the Least Mean Square (LMS) is a well-known adaptive filtering algorithm which plays a significant role in achieving this aim. Nonetheless, the LMS algorithm is usually characterised with low convergence speed in respect to the minimum Mean Square Error (MSE) and flexibility in application. In this paper the Least Mean Square (LMS) algorithm is dealt with a different approach. Contrary to designing LMS filters with fixed step size, variable step size is introduced to improve its convergence speed. An experimental study is considered to formulate a new method for adjusting the step size of the LMS algorithm in this work. Simulation results as well as performance evaluation of the formulated variable step size (VSS-LMS) are presented and compared with the conventional LMS algorithm in terms of MSE and convergence speed.