Speech is the basic way through which information is exchanged among the people. Humans have an audible frequency range of 20 HZ to 20,000HZ and Human speech ranges from 300HZ to 3400 HZ. In this regard one of the most common way that speech is affected is due to background noises such as airport, car, babble, restaurant, street, train station, exhibition noises. As speech is a non-stationary signal which means its frequency or its spectral components changes with respect to time. When the speech and noise signals combines and change continuously to extract the desired speech from it there arises the need for adaptive filters. One such highly popular adaptive filter is Recursive Least square (RLS).Here in this paper a simulink model is designed and implemented for noise cancellation using RLS algorithm and tested for different background noises and the MSE(Mean Square Error) and Signal to Noise Ratio(SNR) are calculated. The speech signals have been taken from NOIZEUS database which contains different signals contaminated with noise. Finally , we can be observe that noise reduction has been achieved audibly.