Performance Analysis of Beamforming Algorithm for Noise Cancellation with Respect to the Arrival Angles of Interference Signal Md.Zahangir Alom Hyo Jong Lee Div. of Computer Science and Engineering.. Div. of Computer Science and Engineering. Chonbuk National University, Jeonju, Korea. Center for Advanced Image and Information Technology. e-mail:zhangiralam@yahoo.com Chonbuk National University, Jeonju, Korea e-mail: hlee@chonbuk.ac.kr Abstract--This article presents performance analysis of multiple channel noise cancellation by beamforming algorithm depending on the variation of arrival of angles of the interference signal. In this article we examined that the effectiveness of the desired signal by the different arrival of angles of interference signals. The variation of the arrival angles of the interference signal considered to with intervals and the desired signal arrival angle is . The desired signal and the optimized multi-layer perception (MLP) were used to calculate the correlation values at different angle. We also used the Backpropagation algorithm as learning rule for MLP and improving our signal quality. This involves a desired signal whilst removing any noise or interference signals which may come from different sources [1][2]. 0 150  0 150  0 30 0 0 Keywords: DSPs,  LMS, MLP, SS, BSS, Nyquist Sampling Rate, BSS. I. INTRODUCTION In the past few years, there has been extensive effort to improve the noise reduction performance for noisy signal. In recent years, scientific literature has contained considerable information about spectral subtraction (SS), Wiener filter, Kalman filter and adaptive filter techniques. Noise reduction in this context means the improvement of the quality or intelligibility of the signal by reducing or removing background interference, signal distortion, and hence, the improvement of the signal to noise ratio(SNR) of the contaminated signal, where the signal and ambient noise are recorded using microphone. Noise reduction techniques such as spectral subtraction, the  LMS algorithm are applied and the listener hears only the signal [11]. One of the important assumptions of this technique is that the listener is acoustically isolated from the environment. In this research Beamforming technique is used for noise cancellation and performance is analyzed by varying hidden layers and number of iterations (epoch). Backpropagation algorithm is used to improve the signal quality and to train the data [8]. Hamming windowing function is used to shortly represent the signals [1][2]. II. ARCHITECTURE OF THE PROPOSED SYSTEM The arrangement for the neural network based on beamforming is depicted in Fig.1. From the Fig. 1 it is clear that there will be a neuron for each array element. A beamforming is an array of sensors connected to the MLP inputs. The electromagnetic signal impinges each antenna at different time [8][9]. AE 1 AE 2 AE N Output Fig. 1: Neural Network for beamforming This time difference is effectively a delay on the signal and the physical spacing of the elements antenna is used to calculate the required delays. In this case the linear array will consist of 6 elements which process the desired signal which has been contaminated with noise. A. Multilayer feedforward neural network Input Hidden Output Fig. 2: Three layer MLP Fig.2 represents the fully connected feedforward three layers MLP. Fig.2 is known as a 7-4-3 architecture because it has 7 input nodes, 4 hidden nodes, and 3 output nodes. In general the input layer has not performed any computation just act as a buffer. The neural network architecture will be optimized by trial and error [9].In this system, 6 0 180 W2i,j - W1i,j - Multilayer feedforward neural network 2010 International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery 978-0-7695-4235-5/10 $26.00 © 2010 IEEE DOI 10.1109/CyberC.2010.78 396