This paper presents the use of Error Back Propagation Algorithm for Adaptive Filter Applications. The Artificial Neural Network (ANN) exploits correlation between the pure speech signal and echo corrupted signal, to generate an estimate of the echo, which in turn subtracts the noise from the echo corrupted signal. Training of ANN is performed using the offline recordings of male and female voice data. Based on the analysis of the recovered pure signal, plots of Signal to Noise Ratio (SNR) vs. Error, Tracking of Weights, Tracking of Echo, Tracking of Error and Variance of Echo with relation to Error has also been shown.